API tools faq
paste
Advertisement
Guest User

Untitled

a guest
Jul 23rd, 2019
129
0
Never
Add comment
Not a member of Pastebin yet? Sign Up, it unlocks many cool features!
text 45.30 KB | None | 0 0
raw download clone embed print report
  1. name: "Openpose"
  2. input: "image"
  3. input_dim: 1
  4. input_dim: 3
  5. input_dim: 1 # This value will be defined at runtime
  6. input_dim: 1 # This value will be defined at runtime
  7. layer {
  8. name: "conv1_1"
  9. type: "Convolution"
  10. bottom: "image"
  11. top: "conv1_1"
  12. param {
  13. lr_mult: 1.0
  14. decay_mult: 1
  15. }
  16. param {
  17. lr_mult: 2.0
  18. decay_mult: 0
  19. }
  20. convolution_param {
  21. num_output: 64
  22. pad: 1
  23. kernel_size: 3
  24. weight_filler {
  25. type: "gaussian"
  26. std: 0.01
  27. }
  28. bias_filler {
  29. type: "constant"
  30. }
  31. }
  32. }
  33. layer {
  34. name: "relu1_1"
  35. type: "ReLU"
  36. bottom: "conv1_1"
  37. top: "conv1_1"
  38. }
  39. layer {
  40. name: "conv1_2"
  41. type: "Convolution"
  42. bottom: "conv1_1"
  43. top: "conv1_2"
  44. param {
  45. lr_mult: 1.0
  46. decay_mult: 1
  47. }
  48. param {
  49. lr_mult: 2.0
  50. decay_mult: 0
  51. }
  52. convolution_param {
  53. num_output: 64
  54. pad: 1
  55. kernel_size: 3
  56. weight_filler {
  57. type: "gaussian"
  58. std: 0.01
  59. }
  60. bias_filler {
  61. type: "constant"
  62. }
  63. }
  64. }
  65. layer {
  66. name: "relu1_2"
  67. type: "ReLU"
  68. bottom: "conv1_2"
  69. top: "conv1_2"
  70. }
  71. layer {
  72. name: "pool1_stage1"
  73. type: "Pooling"
  74. bottom: "conv1_2"
  75. top: "pool1_stage1"
  76. pooling_param {
  77. pool: MAX
  78. kernel_size: 2
  79. stride: 2
  80. }
  81. }
  82. layer {
  83. name: "conv2_1"
  84. type: "Convolution"
  85. bottom: "pool1_stage1"
  86. top: "conv2_1"
  87. param {
  88. lr_mult: 1.0
  89. decay_mult: 1
  90. }
  91. param {
  92. lr_mult: 2.0
  93. decay_mult: 0
  94. }
  95. convolution_param {
  96. num_output: 128
  97. pad: 1
  98. kernel_size: 3
  99. weight_filler {
  100. type: "gaussian"
  101. std: 0.01
  102. }
  103. bias_filler {
  104. type: "constant"
  105. }
  106. }
  107. }
  108. layer {
  109. name: "relu2_1"
  110. type: "ReLU"
  111. bottom: "conv2_1"
  112. top: "conv2_1"
  113. }
  114. layer {
  115. name: "conv2_2"
  116. type: "Convolution"
  117. bottom: "conv2_1"
  118. top: "conv2_2"
  119. param {
  120. lr_mult: 1.0
  121. decay_mult: 1
  122. }
  123. param {
  124. lr_mult: 2.0
  125. decay_mult: 0
  126. }
  127. convolution_param {
  128. num_output: 128
  129. pad: 1
  130. kernel_size: 3
  131. weight_filler {
  132. type: "gaussian"
  133. std: 0.01
  134. }
  135. bias_filler {
  136. type: "constant"
  137. }
  138. }
  139. }
  140. layer {
  141. name: "relu2_2"
  142. type: "ReLU"
  143. bottom: "conv2_2"
  144. top: "conv2_2"
  145. }
  146. layer {
  147. name: "pool2_stage1"
  148. type: "Pooling"
  149. bottom: "conv2_2"
  150. top: "pool2_stage1"
  151. pooling_param {
  152. pool: MAX
  153. kernel_size: 2
  154. stride: 2
  155. }
  156. }
  157. layer {
  158. name: "conv3_1"
  159. type: "Convolution"
  160. bottom: "pool2_stage1"
  161. top: "conv3_1"
  162. param {
  163. lr_mult: 1.0
  164. decay_mult: 1
  165. }
  166. param {
  167. lr_mult: 2.0
  168. decay_mult: 0
  169. }
  170. convolution_param {
  171. num_output: 256
  172. pad: 1
  173. kernel_size: 3
  174. weight_filler {
  175. type: "gaussian"
  176. std: 0.01
  177. }
  178. bias_filler {
  179. type: "constant"
  180. }
  181. }
  182. }
  183. layer {
  184. name: "relu3_1"
  185. type: "ReLU"
  186. bottom: "conv3_1"
  187. top: "conv3_1"
  188. }
  189. layer {
  190. name: "conv3_2"
  191. type: "Convolution"
  192. bottom: "conv3_1"
  193. top: "conv3_2"
  194. param {
  195. lr_mult: 1.0
  196. decay_mult: 1
  197. }
  198. param {
  199. lr_mult: 2.0
  200. decay_mult: 0
  201. }
  202. convolution_param {
  203. num_output: 256
  204. pad: 1
  205. kernel_size: 3
  206. weight_filler {
  207. type: "gaussian"
  208. std: 0.01
  209. }
  210. bias_filler {
  211. type: "constant"
  212. }
  213. }
  214. }
  215. layer {
  216. name: "relu3_2"
  217. type: "ReLU"
  218. bottom: "conv3_2"
  219. top: "conv3_2"
  220. }
  221. layer {
  222. name: "conv3_3"
  223. type: "Convolution"
  224. bottom: "conv3_2"
  225. top: "conv3_3"
  226. param {
  227. lr_mult: 1.0
  228. decay_mult: 1
  229. }
  230. param {
  231. lr_mult: 2.0
  232. decay_mult: 0
  233. }
  234. convolution_param {
  235. num_output: 256
  236. pad: 1
  237. kernel_size: 3
  238. weight_filler {
  239. type: "gaussian"
  240. std: 0.01
  241. }
  242. bias_filler {
  243. type: "constant"
  244. }
  245. }
  246. }
  247. layer {
  248. name: "relu3_3"
  249. type: "ReLU"
  250. bottom: "conv3_3"
  251. top: "conv3_3"
  252. }
  253. layer {
  254. name: "conv3_4"
  255. type: "Convolution"
  256. bottom: "conv3_3"
  257. top: "conv3_4"
  258. param {
  259. lr_mult: 1.0
  260. decay_mult: 1
  261. }
  262. param {
  263. lr_mult: 2.0
  264. decay_mult: 0
  265. }
  266. convolution_param {
  267. num_output: 256
  268. pad: 1
  269. kernel_size: 3
  270. weight_filler {
  271. type: "gaussian"
  272. std: 0.01
  273. }
  274. bias_filler {
  275. type: "constant"
  276. }
  277. }
  278. }
  279. layer {
  280. name: "relu3_4"
  281. type: "ReLU"
  282. bottom: "conv3_4"
  283. top: "conv3_4"
  284. }
  285. layer {
  286. name: "pool3_stage1"
  287. type: "Pooling"
  288. bottom: "conv3_4"
  289. top: "pool3_stage1"
  290. pooling_param {
  291. pool: MAX
  292. kernel_size: 2
  293. stride: 2
  294. }
  295. }
  296. layer {
  297. name: "conv4_1"
  298. type: "Convolution"
  299. bottom: "pool3_stage1"
  300. top: "conv4_1"
  301. param {
  302. lr_mult: 1.0
  303. decay_mult: 1
  304. }
  305. param {
  306. lr_mult: 2.0
  307. decay_mult: 0
  308. }
  309. convolution_param {
  310. num_output: 512
  311. pad: 1
  312. kernel_size: 3
  313. weight_filler {
  314. type: "gaussian"
  315. std: 0.01
  316. }
  317. bias_filler {
  318. type: "constant"
  319. }
  320. }
  321. }
  322. layer {
  323. name: "relu4_1"
  324. type: "ReLU"
  325. bottom: "conv4_1"
  326. top: "conv4_1"
  327. }
  328. layer {
  329. name: "conv4_2"
  330. type: "Convolution"
  331. bottom: "conv4_1"
  332. top: "conv4_2"
  333. param {
  334. lr_mult: 1.0
  335. decay_mult: 1
  336. }
  337. param {
  338. lr_mult: 2.0
  339. decay_mult: 0
  340. }
  341. convolution_param {
  342. num_output: 512
  343. pad: 1
  344. kernel_size: 3
  345. weight_filler {
  346. type: "gaussian"
  347. std: 0.01
  348. }
  349. bias_filler {
  350. type: "constant"
  351. }
  352. }
  353. }
  354. layer {
  355. name: "relu4_2"
  356. type: "ReLU"
  357. bottom: "conv4_2"
  358. top: "conv4_2"
  359. }
  360. layer {
  361. name: "conv4_3_CPM"
  362. type: "Convolution"
  363. bottom: "conv4_2"
  364. top: "conv4_3_CPM"
  365. param {
  366. lr_mult: 1.0
  367. decay_mult: 1
  368. }
  369. param {
  370. lr_mult: 2.0
  371. decay_mult: 0
  372. }
  373. convolution_param {
  374. num_output: 256
  375. pad: 1
  376. kernel_size: 3
  377. weight_filler {
  378. type: "gaussian"
  379. std: 0.01
  380. }
  381. bias_filler {
  382. type: "constant"
  383. }
  384. }
  385. }
  386. layer {
  387. name: "relu4_3_CPM"
  388. type: "ReLU"
  389. bottom: "conv4_3_CPM"
  390. top: "conv4_3_CPM"
  391. }
  392. layer {
  393. name: "conv4_4_CPM"
  394. type: "Convolution"
  395. bottom: "conv4_3_CPM"
  396. top: "conv4_4_CPM"
  397. param {
  398. lr_mult: 1.0
  399. decay_mult: 1
  400. }
  401. param {
  402. lr_mult: 2.0
  403. decay_mult: 0
  404. }
  405. convolution_param {
  406. num_output: 128
  407. pad: 1
  408. kernel_size: 3
  409. weight_filler {
  410. type: "gaussian"
  411. std: 0.01
  412. }
  413. bias_filler {
  414. type: "constant"
  415. }
  416. }
  417. }
  418. layer {
  419. name: "relu4_4_CPM"
  420. type: "ReLU"
  421. bottom: "conv4_4_CPM"
  422. top: "conv4_4_CPM"
  423. }
  424. layer {
  425. name: "conv5_1_CPM_L1"
  426. type: "Convolution"
  427. bottom: "conv4_4_CPM"
  428. top: "conv5_1_CPM_L1"
  429. param {
  430. lr_mult: 1.0
  431. decay_mult: 1
  432. }
  433. param {
  434. lr_mult: 2.0
  435. decay_mult: 0
  436. }
  437. convolution_param {
  438. num_output: 128
  439. pad: 1
  440. kernel_size: 3
  441. weight_filler {
  442. type: "gaussian"
  443. std: 0.01
  444. }
  445. bias_filler {
  446. type: "constant"
  447. }
  448. }
  449. }
  450. layer {
  451. name: "relu5_1_CPM_L1"
  452. type: "ReLU"
  453. bottom: "conv5_1_CPM_L1"
  454. top: "conv5_1_CPM_L1"
  455. }
  456. layer {
  457. name: "conv5_1_CPM_L2"
  458. type: "Convolution"
  459. bottom: "conv4_4_CPM"
  460. top: "conv5_1_CPM_L2"
  461. param {
  462. lr_mult: 1.0
  463. decay_mult: 1
  464. }
  465. param {
  466. lr_mult: 2.0
  467. decay_mult: 0
  468. }
  469. convolution_param {
  470. num_output: 128
  471. pad: 1
  472. kernel_size: 3
  473. weight_filler {
  474. type: "gaussian"
  475. std: 0.01
  476. }
  477. bias_filler {
  478. type: "constant"
  479. }
  480. }
  481. }
  482. layer {
  483. name: "relu5_1_CPM_L2"
  484. type: "ReLU"
  485. bottom: "conv5_1_CPM_L2"
  486. top: "conv5_1_CPM_L2"
  487. }
  488. layer {
  489. name: "conv5_2_CPM_L1"
  490. type: "Convolution"
  491. bottom: "conv5_1_CPM_L1"
  492. top: "conv5_2_CPM_L1"
  493. param {
  494. lr_mult: 1.0
  495. decay_mult: 1
  496. }
  497. param {
  498. lr_mult: 2.0
  499. decay_mult: 0
  500. }
  501. convolution_param {
  502. num_output: 128
  503. pad: 1
  504. kernel_size: 3
  505. weight_filler {
  506. type: "gaussian"
  507. std: 0.01
  508. }
  509. bias_filler {
  510. type: "constant"
  511. }
  512. }
  513. }
  514. layer {
  515. name: "relu5_2_CPM_L1"
  516. type: "ReLU"
  517. bottom: "conv5_2_CPM_L1"
  518. top: "conv5_2_CPM_L1"
  519. }
  520. layer {
  521. name: "conv5_2_CPM_L2"
  522. type: "Convolution"
  523. bottom: "conv5_1_CPM_L2"
  524. top: "conv5_2_CPM_L2"
  525. param {
  526. lr_mult: 1.0
  527. decay_mult: 1
  528. }
  529. param {
  530. lr_mult: 2.0
  531. decay_mult: 0
  532. }
  533. convolution_param {
  534. num_output: 128
  535. pad: 1
  536. kernel_size: 3
  537. weight_filler {
  538. type: "gaussian"
  539. std: 0.01
  540. }
  541. bias_filler {
  542. type: "constant"
  543. }
  544. }
  545. }
  546. layer {
  547. name: "relu5_2_CPM_L2"
  548. type: "ReLU"
  549. bottom: "conv5_2_CPM_L2"
  550. top: "conv5_2_CPM_L2"
  551. }
  552. layer {
  553. name: "conv5_3_CPM_L1"
  554. type: "Convolution"
  555. bottom: "conv5_2_CPM_L1"
  556. top: "conv5_3_CPM_L1"
  557. param {
  558. lr_mult: 1.0
  559. decay_mult: 1
  560. }
  561. param {
  562. lr_mult: 2.0
  563. decay_mult: 0
  564. }
  565. convolution_param {
  566. num_output: 128
  567. pad: 1
  568. kernel_size: 3
  569. weight_filler {
  570. type: "gaussian"
  571. std: 0.01
  572. }
  573. bias_filler {
  574. type: "constant"
  575. }
  576. }
  577. }
  578. layer {
  579. name: "relu5_3_CPM_L1"
  580. type: "ReLU"
  581. bottom: "conv5_3_CPM_L1"
  582. top: "conv5_3_CPM_L1"
  583. }
  584. layer {
  585. name: "conv5_3_CPM_L2"
  586. type: "Convolution"
  587. bottom: "conv5_2_CPM_L2"
  588. top: "conv5_3_CPM_L2"
  589. param {
  590. lr_mult: 1.0
  591. decay_mult: 1
  592. }
  593. param {
  594. lr_mult: 2.0
  595. decay_mult: 0
  596. }
  597. convolution_param {
  598. num_output: 128
  599. pad: 1
  600. kernel_size: 3
  601. weight_filler {
  602. type: "gaussian"
  603. std: 0.01
  604. }
  605. bias_filler {
  606. type: "constant"
  607. }
  608. }
  609. }
  610. layer {
  611. name: "relu5_3_CPM_L2"
  612. type: "ReLU"
  613. bottom: "conv5_3_CPM_L2"
  614. top: "conv5_3_CPM_L2"
  615. }
  616. layer {
  617. name: "conv5_4_CPM_L1"
  618. type: "Convolution"
  619. bottom: "conv5_3_CPM_L1"
  620. top: "conv5_4_CPM_L1"
  621. param {
  622. lr_mult: 1.0
  623. decay_mult: 1
  624. }
  625. param {
  626. lr_mult: 2.0
  627. decay_mult: 0
  628. }
  629. convolution_param {
  630. num_output: 512
  631. pad: 0
  632. kernel_size: 1
  633. weight_filler {
  634. type: "gaussian"
  635. std: 0.01
  636. }
  637. bias_filler {
  638. type: "constant"
  639. }
  640. }
  641. }
  642. layer {
  643. name: "relu5_4_CPM_L1"
  644. type: "ReLU"
  645. bottom: "conv5_4_CPM_L1"
  646. top: "conv5_4_CPM_L1"
  647. }
  648. layer {
  649. name: "conv5_4_CPM_L2"
  650. type: "Convolution"
  651. bottom: "conv5_3_CPM_L2"
  652. top: "conv5_4_CPM_L2"
  653. param {
  654. lr_mult: 1.0
  655. decay_mult: 1
  656. }
  657. param {
  658. lr_mult: 2.0
  659. decay_mult: 0
  660. }
  661. convolution_param {
  662. num_output: 512
  663. pad: 0
  664. kernel_size: 1
  665. weight_filler {
  666. type: "gaussian"
  667. std: 0.01
  668. }
  669. bias_filler {
  670. type: "constant"
  671. }
  672. }
  673. }
  674. layer {
  675. name: "relu5_4_CPM_L2"
  676. type: "ReLU"
  677. bottom: "conv5_4_CPM_L2"
  678. top: "conv5_4_CPM_L2"
  679. }
  680. layer {
  681. name: "conv5_5_CPM_L1"
  682. type: "Convolution"
  683. bottom: "conv5_4_CPM_L1"
  684. top: "conv5_5_CPM_L1"
  685. param {
  686. lr_mult: 1.0
  687. decay_mult: 1
  688. }
  689. param {
  690. lr_mult: 2.0
  691. decay_mult: 0
  692. }
  693. convolution_param {
  694. num_output: 38
  695. pad: 0
  696. kernel_size: 1
  697. weight_filler {
  698. type: "gaussian"
  699. std: 0.01
  700. }
  701. bias_filler {
  702. type: "constant"
  703. }
  704. }
  705. }
  706. layer {
  707. name: "conv5_5_CPM_L2"
  708. type: "Convolution"
  709. bottom: "conv5_4_CPM_L2"
  710. top: "conv5_5_CPM_L2"
  711. param {
  712. lr_mult: 1.0
  713. decay_mult: 1
  714. }
  715. param {
  716. lr_mult: 2.0
  717. decay_mult: 0
  718. }
  719. convolution_param {
  720. num_output: 19
  721. pad: 0
  722. kernel_size: 1
  723. weight_filler {
  724. type: "gaussian"
  725. std: 0.01
  726. }
  727. bias_filler {
  728. type: "constant"
  729. }
  730. }
  731. }
  732. layer {
  733. name: "concat_stage2"
  734. type: "Concat"
  735. bottom: "conv5_5_CPM_L1"
  736. bottom: "conv5_5_CPM_L2"
  737. bottom: "conv4_4_CPM"
  738. top: "concat_stage2"
  739. concat_param {
  740. axis: 1
  741. }
  742. }
  743. layer {
  744. name: "Mconv1_stage2_L1"
  745. type: "Convolution"
  746. bottom: "concat_stage2"
  747. top: "Mconv1_stage2_L1"
  748. param {
  749. lr_mult: 4.0
  750. decay_mult: 1
  751. }
  752. param {
  753. lr_mult: 8.0
  754. decay_mult: 0
  755. }
  756. convolution_param {
  757. num_output: 128
  758. pad: 3
  759. kernel_size: 7
  760. weight_filler {
  761. type: "gaussian"
  762. std: 0.01
  763. }
  764. bias_filler {
  765. type: "constant"
  766. }
  767. }
  768. }
  769. layer {
  770. name: "Mrelu1_stage2_L1"
  771. type: "ReLU"
  772. bottom: "Mconv1_stage2_L1"
  773. top: "Mconv1_stage2_L1"
  774. }
  775. layer {
  776. name: "Mconv1_stage2_L2"
  777. type: "Convolution"
  778. bottom: "concat_stage2"
  779. top: "Mconv1_stage2_L2"
  780. param {
  781. lr_mult: 4.0
  782. decay_mult: 1
  783. }
  784. param {
  785. lr_mult: 8.0
  786. decay_mult: 0
  787. }
  788. convolution_param {
  789. num_output: 128
  790. pad: 3
  791. kernel_size: 7
  792. weight_filler {
  793. type: "gaussian"
  794. std: 0.01
  795. }
  796. bias_filler {
  797. type: "constant"
  798. }
  799. }
  800. }
  801. layer {
  802. name: "Mrelu1_stage2_L2"
  803. type: "ReLU"
  804. bottom: "Mconv1_stage2_L2"
  805. top: "Mconv1_stage2_L2"
  806. }
  807. layer {
  808. name: "Mconv2_stage2_L1"
  809. type: "Convolution"
  810. bottom: "Mconv1_stage2_L1"
  811. top: "Mconv2_stage2_L1"
  812. param {
  813. lr_mult: 4.0
  814. decay_mult: 1
  815. }
  816. param {
  817. lr_mult: 8.0
  818. decay_mult: 0
  819. }
  820. convolution_param {
  821. num_output: 128
  822. pad: 3
  823. kernel_size: 7
  824. weight_filler {
  825. type: "gaussian"
  826. std: 0.01
  827. }
  828. bias_filler {
  829. type: "constant"
  830. }
  831. }
  832. }
  833. layer {
  834. name: "Mrelu2_stage2_L1"
  835. type: "ReLU"
  836. bottom: "Mconv2_stage2_L1"
  837. top: "Mconv2_stage2_L1"
  838. }
  839. layer {
  840. name: "Mconv2_stage2_L2"
  841. type: "Convolution"
  842. bottom: "Mconv1_stage2_L2"
  843. top: "Mconv2_stage2_L2"
  844. param {
  845. lr_mult: 4.0
  846. decay_mult: 1
  847. }
  848. param {
  849. lr_mult: 8.0
  850. decay_mult: 0
  851. }
  852. convolution_param {
  853. num_output: 128
  854. pad: 3
  855. kernel_size: 7
  856. weight_filler {
  857. type: "gaussian"
  858. std: 0.01
  859. }
  860. bias_filler {
  861. type: "constant"
  862. }
  863. }
  864. }
  865. layer {
  866. name: "Mrelu2_stage2_L2"
  867. type: "ReLU"
  868. bottom: "Mconv2_stage2_L2"
  869. top: "Mconv2_stage2_L2"
  870. }
  871. layer {
  872. name: "Mconv3_stage2_L1"
  873. type: "Convolution"
  874. bottom: "Mconv2_stage2_L1"
  875. top: "Mconv3_stage2_L1"
  876. param {
  877. lr_mult: 4.0
  878. decay_mult: 1
  879. }
  880. param {
  881. lr_mult: 8.0
  882. decay_mult: 0
  883. }
  884. convolution_param {
  885. num_output: 128
  886. pad: 3
  887. kernel_size: 7
  888. weight_filler {
  889. type: "gaussian"
  890. std: 0.01
  891. }
  892. bias_filler {
  893. type: "constant"
  894. }
  895. }
  896. }
  897. layer {
  898. name: "Mrelu3_stage2_L1"
  899. type: "ReLU"
  900. bottom: "Mconv3_stage2_L1"
  901. top: "Mconv3_stage2_L1"
  902. }
  903. layer {
  904. name: "Mconv3_stage2_L2"
  905. type: "Convolution"
  906. bottom: "Mconv2_stage2_L2"
  907. top: "Mconv3_stage2_L2"
  908. param {
  909. lr_mult: 4.0
  910. decay_mult: 1
  911. }
  912. param {
  913. lr_mult: 8.0
  914. decay_mult: 0
  915. }
  916. convolution_param {
  917. num_output: 128
  918. pad: 3
  919. kernel_size: 7
  920. weight_filler {
  921. type: "gaussian"
  922. std: 0.01
  923. }
  924. bias_filler {
  925. type: "constant"
  926. }
  927. }
  928. }
  929. layer {
  930. name: "Mrelu3_stage2_L2"
  931. type: "ReLU"
  932. bottom: "Mconv3_stage2_L2"
  933. top: "Mconv3_stage2_L2"
  934. }
  935. layer {
  936. name: "Mconv4_stage2_L1"
  937. type: "Convolution"
  938. bottom: "Mconv3_stage2_L1"
  939. top: "Mconv4_stage2_L1"
  940. param {
  941. lr_mult: 4.0
  942. decay_mult: 1
  943. }
  944. param {
  945. lr_mult: 8.0
  946. decay_mult: 0
  947. }
  948. convolution_param {
  949. num_output: 128
  950. pad: 3
  951. kernel_size: 7
  952. weight_filler {
  953. type: "gaussian"
  954. std: 0.01
  955. }
  956. bias_filler {
  957. type: "constant"
  958. }
  959. }
  960. }
  961. layer {
  962. name: "Mrelu4_stage2_L1"
  963. type: "ReLU"
  964. bottom: "Mconv4_stage2_L1"
  965. top: "Mconv4_stage2_L1"
  966. }
  967. layer {
  968. name: "Mconv4_stage2_L2"
  969. type: "Convolution"
  970. bottom: "Mconv3_stage2_L2"
  971. top: "Mconv4_stage2_L2"
  972. param {
  973. lr_mult: 4.0
  974. decay_mult: 1
  975. }
  976. param {
  977. lr_mult: 8.0
  978. decay_mult: 0
  979. }
  980. convolution_param {
  981. num_output: 128
  982. pad: 3
  983. kernel_size: 7
  984. weight_filler {
  985. type: "gaussian"
  986. std: 0.01
  987. }
  988. bias_filler {
  989. type: "constant"
  990. }
  991. }
  992. }
  993. layer {
  994. name: "Mrelu4_stage2_L2"
  995. type: "ReLU"
  996. bottom: "Mconv4_stage2_L2"
  997. top: "Mconv4_stage2_L2"
  998. }
  999. layer {
  1000. name: "Mconv5_stage2_L1"
  1001. type: "Convolution"
  1002. bottom: "Mconv4_stage2_L1"
  1003. top: "Mconv5_stage2_L1"
  1004. param {
  1005. lr_mult: 4.0
  1006. decay_mult: 1
  1007. }
  1008. param {
  1009. lr_mult: 8.0
  1010. decay_mult: 0
  1011. }
  1012. convolution_param {
  1013. num_output: 128
  1014. pad: 3
  1015. kernel_size: 7
  1016. weight_filler {
  1017. type: "gaussian"
  1018. std: 0.01
  1019. }
  1020. bias_filler {
  1021. type: "constant"
  1022. }
  1023. }
  1024. }
  1025. layer {
  1026. name: "Mrelu5_stage2_L1"
  1027. type: "ReLU"
  1028. bottom: "Mconv5_stage2_L1"
  1029. top: "Mconv5_stage2_L1"
  1030. }
  1031. layer {
  1032. name: "Mconv5_stage2_L2"
  1033. type: "Convolution"
  1034. bottom: "Mconv4_stage2_L2"
  1035. top: "Mconv5_stage2_L2"
  1036. param {
  1037. lr_mult: 4.0
  1038. decay_mult: 1
  1039. }
  1040. param {
  1041. lr_mult: 8.0
  1042. decay_mult: 0
  1043. }
  1044. convolution_param {
  1045. num_output: 128
  1046. pad: 3
  1047. kernel_size: 7
  1048. weight_filler {
  1049. type: "gaussian"
  1050. std: 0.01
  1051. }
  1052. bias_filler {
  1053. type: "constant"
  1054. }
  1055. }
  1056. }
  1057. layer {
  1058. name: "Mrelu5_stage2_L2"
  1059. type: "ReLU"
  1060. bottom: "Mconv5_stage2_L2"
  1061. top: "Mconv5_stage2_L2"
  1062. }
  1063. layer {
  1064. name: "Mconv6_stage2_L1"
  1065. type: "Convolution"
  1066. bottom: "Mconv5_stage2_L1"
  1067. top: "Mconv6_stage2_L1"
  1068. param {
  1069. lr_mult: 4.0
  1070. decay_mult: 1
  1071. }
  1072. param {
  1073. lr_mult: 8.0
  1074. decay_mult: 0
  1075. }
  1076. convolution_param {
  1077. num_output: 128
  1078. pad: 0
  1079. kernel_size: 1
  1080. weight_filler {
  1081. type: "gaussian"
  1082. std: 0.01
  1083. }
  1084. bias_filler {
  1085. type: "constant"
  1086. }
  1087. }
  1088. }
  1089. layer {
  1090. name: "Mrelu6_stage2_L1"
  1091. type: "ReLU"
  1092. bottom: "Mconv6_stage2_L1"
  1093. top: "Mconv6_stage2_L1"
  1094. }
  1095. layer {
  1096. name: "Mconv6_stage2_L2"
  1097. type: "Convolution"
  1098. bottom: "Mconv5_stage2_L2"
  1099. top: "Mconv6_stage2_L2"
  1100. param {
  1101. lr_mult: 4.0
  1102. decay_mult: 1
  1103. }
  1104. param {
  1105. lr_mult: 8.0
  1106. decay_mult: 0
  1107. }
  1108. convolution_param {
  1109. num_output: 128
  1110. pad: 0
  1111. kernel_size: 1
  1112. weight_filler {
  1113. type: "gaussian"
  1114. std: 0.01
  1115. }
  1116. bias_filler {
  1117. type: "constant"
  1118. }
  1119. }
  1120. }
  1121. layer {
  1122. name: "Mrelu6_stage2_L2"
  1123. type: "ReLU"
  1124. bottom: "Mconv6_stage2_L2"
  1125. top: "Mconv6_stage2_L2"
  1126. }
  1127. layer {
  1128. name: "Mconv7_stage2_L1"
  1129. type: "Convolution"
  1130. bottom: "Mconv6_stage2_L1"
  1131. top: "Mconv7_stage2_L1"
  1132. param {
  1133. lr_mult: 4.0
  1134. decay_mult: 1
  1135. }
  1136. param {
  1137. lr_mult: 8.0
  1138. decay_mult: 0
  1139. }
  1140. convolution_param {
  1141. num_output: 38
  1142. pad: 0
  1143. kernel_size: 1
  1144. weight_filler {
  1145. type: "gaussian"
  1146. std: 0.01
  1147. }
  1148. bias_filler {
  1149. type: "constant"
  1150. }
  1151. }
  1152. }
  1153. layer {
  1154. name: "Mconv7_stage2_L2"
  1155. type: "Convolution"
  1156. bottom: "Mconv6_stage2_L2"
  1157. top: "Mconv7_stage2_L2"
  1158. param {
  1159. lr_mult: 4.0
  1160. decay_mult: 1
  1161. }
  1162. param {
  1163. lr_mult: 8.0
  1164. decay_mult: 0
  1165. }
  1166. convolution_param {
  1167. num_output: 19
  1168. pad: 0
  1169. kernel_size: 1
  1170. weight_filler {
  1171. type: "gaussian"
  1172. std: 0.01
  1173. }
  1174. bias_filler {
  1175. type: "constant"
  1176. }
  1177. }
  1178. }
  1179. layer {
  1180. name: "concat_stage3"
  1181. type: "Concat"
  1182. bottom: "Mconv7_stage2_L1"
  1183. bottom: "Mconv7_stage2_L2"
  1184. bottom: "conv4_4_CPM"
  1185. top: "concat_stage3"
  1186. concat_param {
  1187. axis: 1
  1188. }
  1189. }
  1190. layer {
  1191. name: "Mconv1_stage3_L1"
  1192. type: "Convolution"
  1193. bottom: "concat_stage3"
  1194. top: "Mconv1_stage3_L1"
  1195. param {
  1196. lr_mult: 4.0
  1197. decay_mult: 1
  1198. }
  1199. param {
  1200. lr_mult: 8.0
  1201. decay_mult: 0
  1202. }
  1203. convolution_param {
  1204. num_output: 128
  1205. pad: 3
  1206. kernel_size: 7
  1207. weight_filler {
  1208. type: "gaussian"
  1209. std: 0.01
  1210. }
  1211. bias_filler {
  1212. type: "constant"
  1213. }
  1214. }
  1215. }
  1216. layer {
  1217. name: "Mrelu1_stage3_L1"
  1218. type: "ReLU"
  1219. bottom: "Mconv1_stage3_L1"
  1220. top: "Mconv1_stage3_L1"
  1221. }
  1222. layer {
  1223. name: "Mconv1_stage3_L2"
  1224. type: "Convolution"
  1225. bottom: "concat_stage3"
  1226. top: "Mconv1_stage3_L2"
  1227. param {
  1228. lr_mult: 4.0
  1229. decay_mult: 1
  1230. }
  1231. param {
  1232. lr_mult: 8.0
  1233. decay_mult: 0
  1234. }
  1235. convolution_param {
  1236. num_output: 128
  1237. pad: 3
  1238. kernel_size: 7
  1239. weight_filler {
  1240. type: "gaussian"
  1241. std: 0.01
  1242. }
  1243. bias_filler {
  1244. type: "constant"
  1245. }
  1246. }
  1247. }
  1248. layer {
  1249. name: "Mrelu1_stage3_L2"
  1250. type: "ReLU"
  1251. bottom: "Mconv1_stage3_L2"
  1252. top: "Mconv1_stage3_L2"
  1253. }
  1254. layer {
  1255. name: "Mconv2_stage3_L1"
  1256. type: "Convolution"
  1257. bottom: "Mconv1_stage3_L1"
  1258. top: "Mconv2_stage3_L1"
  1259. param {
  1260. lr_mult: 4.0
  1261. decay_mult: 1
  1262. }
  1263. param {
  1264. lr_mult: 8.0
  1265. decay_mult: 0
  1266. }
  1267. convolution_param {
  1268. num_output: 128
  1269. pad: 3
  1270. kernel_size: 7
  1271. weight_filler {
  1272. type: "gaussian"
  1273. std: 0.01
  1274. }
  1275. bias_filler {
  1276. type: "constant"
  1277. }
  1278. }
  1279. }
  1280. layer {
  1281. name: "Mrelu2_stage3_L1"
  1282. type: "ReLU"
  1283. bottom: "Mconv2_stage3_L1"
  1284. top: "Mconv2_stage3_L1"
  1285. }
  1286. layer {
  1287. name: "Mconv2_stage3_L2"
  1288. type: "Convolution"
  1289. bottom: "Mconv1_stage3_L2"
  1290. top: "Mconv2_stage3_L2"
  1291. param {
  1292. lr_mult: 4.0
  1293. decay_mult: 1
  1294. }
  1295. param {
  1296. lr_mult: 8.0
  1297. decay_mult: 0
  1298. }
  1299. convolution_param {
  1300. num_output: 128
  1301. pad: 3
  1302. kernel_size: 7
  1303. weight_filler {
  1304. type: "gaussian"
  1305. std: 0.01
  1306. }
  1307. bias_filler {
  1308. type: "constant"
  1309. }
  1310. }
  1311. }
  1312. layer {
  1313. name: "Mrelu2_stage3_L2"
  1314. type: "ReLU"
  1315. bottom: "Mconv2_stage3_L2"
  1316. top: "Mconv2_stage3_L2"
  1317. }
  1318. layer {
  1319. name: "Mconv3_stage3_L1"
  1320. type: "Convolution"
  1321. bottom: "Mconv2_stage3_L1"
  1322. top: "Mconv3_stage3_L1"
  1323. param {
  1324. lr_mult: 4.0
  1325. decay_mult: 1
  1326. }
  1327. param {
  1328. lr_mult: 8.0
  1329. decay_mult: 0
  1330. }
  1331. convolution_param {
  1332. num_output: 128
  1333. pad: 3
  1334. kernel_size: 7
  1335. weight_filler {
  1336. type: "gaussian"
  1337. std: 0.01
  1338. }
  1339. bias_filler {
  1340. type: "constant"
  1341. }
  1342. }
  1343. }
  1344. layer {
  1345. name: "Mrelu3_stage3_L1"
  1346. type: "ReLU"
  1347. bottom: "Mconv3_stage3_L1"
  1348. top: "Mconv3_stage3_L1"
  1349. }
  1350. layer {
  1351. name: "Mconv3_stage3_L2"
  1352. type: "Convolution"
  1353. bottom: "Mconv2_stage3_L2"
  1354. top: "Mconv3_stage3_L2"
  1355. param {
  1356. lr_mult: 4.0
  1357. decay_mult: 1
  1358. }
  1359. param {
  1360. lr_mult: 8.0
  1361. decay_mult: 0
  1362. }
  1363. convolution_param {
  1364. num_output: 128
  1365. pad: 3
  1366. kernel_size: 7
  1367. weight_filler {
  1368. type: "gaussian"
  1369. std: 0.01
  1370. }
  1371. bias_filler {
  1372. type: "constant"
  1373. }
  1374. }
  1375. }
  1376. layer {
  1377. name: "Mrelu3_stage3_L2"
  1378. type: "ReLU"
  1379. bottom: "Mconv3_stage3_L2"
  1380. top: "Mconv3_stage3_L2"
  1381. }
  1382. layer {
  1383. name: "Mconv4_stage3_L1"
  1384. type: "Convolution"
  1385. bottom: "Mconv3_stage3_L1"
  1386. top: "Mconv4_stage3_L1"
  1387. param {
  1388. lr_mult: 4.0
  1389. decay_mult: 1
  1390. }
  1391. param {
  1392. lr_mult: 8.0
  1393. decay_mult: 0
  1394. }
  1395. convolution_param {
  1396. num_output: 128
  1397. pad: 3
  1398. kernel_size: 7
  1399. weight_filler {
  1400. type: "gaussian"
  1401. std: 0.01
  1402. }
  1403. bias_filler {
  1404. type: "constant"
  1405. }
  1406. }
  1407. }
  1408. layer {
  1409. name: "Mrelu4_stage3_L1"
  1410. type: "ReLU"
  1411. bottom: "Mconv4_stage3_L1"
  1412. top: "Mconv4_stage3_L1"
  1413. }
  1414. layer {
  1415. name: "Mconv4_stage3_L2"
  1416. type: "Convolution"
  1417. bottom: "Mconv3_stage3_L2"
  1418. top: "Mconv4_stage3_L2"
  1419. param {
  1420. lr_mult: 4.0
  1421. decay_mult: 1
  1422. }
  1423. param {
  1424. lr_mult: 8.0
  1425. decay_mult: 0
  1426. }
  1427. convolution_param {
  1428. num_output: 128
  1429. pad: 3
  1430. kernel_size: 7
  1431. weight_filler {
  1432. type: "gaussian"
  1433. std: 0.01
  1434. }
  1435. bias_filler {
  1436. type: "constant"
  1437. }
  1438. }
  1439. }
  1440. layer {
  1441. name: "Mrelu4_stage3_L2"
  1442. type: "ReLU"
  1443. bottom: "Mconv4_stage3_L2"
  1444. top: "Mconv4_stage3_L2"
  1445. }
  1446. layer {
  1447. name: "Mconv5_stage3_L1"
  1448. type: "Convolution"
  1449. bottom: "Mconv4_stage3_L1"
  1450. top: "Mconv5_stage3_L1"
  1451. param {
  1452. lr_mult: 4.0
  1453. decay_mult: 1
  1454. }
  1455. param {
  1456. lr_mult: 8.0
  1457. decay_mult: 0
  1458. }
  1459. convolution_param {
  1460. num_output: 128
  1461. pad: 3
  1462. kernel_size: 7
  1463. weight_filler {
  1464. type: "gaussian"
  1465. std: 0.01
  1466. }
  1467. bias_filler {
  1468. type: "constant"
  1469. }
  1470. }
  1471. }
  1472. layer {
  1473. name: "Mrelu5_stage3_L1"
  1474. type: "ReLU"
  1475. bottom: "Mconv5_stage3_L1"
  1476. top: "Mconv5_stage3_L1"
  1477. }
  1478. layer {
  1479. name: "Mconv5_stage3_L2"
  1480. type: "Convolution"
  1481. bottom: "Mconv4_stage3_L2"
  1482. top: "Mconv5_stage3_L2"
  1483. param {
  1484. lr_mult: 4.0
  1485. decay_mult: 1
  1486. }
  1487. param {
  1488. lr_mult: 8.0
  1489. decay_mult: 0
  1490. }
  1491. convolution_param {
  1492. num_output: 128
  1493. pad: 3
  1494. kernel_size: 7
  1495. weight_filler {
  1496. type: "gaussian"
  1497. std: 0.01
  1498. }
  1499. bias_filler {
  1500. type: "constant"
  1501. }
  1502. }
  1503. }
  1504. layer {
  1505. name: "Mrelu5_stage3_L2"
  1506. type: "ReLU"
  1507. bottom: "Mconv5_stage3_L2"
  1508. top: "Mconv5_stage3_L2"
  1509. }
  1510. layer {
  1511. name: "Mconv6_stage3_L1"
  1512. type: "Convolution"
  1513. bottom: "Mconv5_stage3_L1"
  1514. top: "Mconv6_stage3_L1"
  1515. param {
  1516. lr_mult: 4.0
  1517. decay_mult: 1
  1518. }
  1519. param {
  1520. lr_mult: 8.0
  1521. decay_mult: 0
  1522. }
  1523. convolution_param {
  1524. num_output: 128
  1525. pad: 0
  1526. kernel_size: 1
  1527. weight_filler {
  1528. type: "gaussian"
  1529. std: 0.01
  1530. }
  1531. bias_filler {
  1532. type: "constant"
  1533. }
  1534. }
  1535. }
  1536. layer {
  1537. name: "Mrelu6_stage3_L1"
  1538. type: "ReLU"
  1539. bottom: "Mconv6_stage3_L1"
  1540. top: "Mconv6_stage3_L1"
  1541. }
  1542. layer {
  1543. name: "Mconv6_stage3_L2"
  1544. type: "Convolution"
  1545. bottom: "Mconv5_stage3_L2"
  1546. top: "Mconv6_stage3_L2"
  1547. param {
  1548. lr_mult: 4.0
  1549. decay_mult: 1
  1550. }
  1551. param {
  1552. lr_mult: 8.0
  1553. decay_mult: 0
  1554. }
  1555. convolution_param {
  1556. num_output: 128
  1557. pad: 0
  1558. kernel_size: 1
  1559. weight_filler {
  1560. type: "gaussian"
  1561. std: 0.01
  1562. }
  1563. bias_filler {
  1564. type: "constant"
  1565. }
  1566. }
  1567. }
  1568. layer {
  1569. name: "Mrelu6_stage3_L2"
  1570. type: "ReLU"
  1571. bottom: "Mconv6_stage3_L2"
  1572. top: "Mconv6_stage3_L2"
  1573. }
  1574. layer {
  1575. name: "Mconv7_stage3_L1"
  1576. type: "Convolution"
  1577. bottom: "Mconv6_stage3_L1"
  1578. top: "Mconv7_stage3_L1"
  1579. param {
  1580. lr_mult: 4.0
  1581. decay_mult: 1
  1582. }
  1583. param {
  1584. lr_mult: 8.0
  1585. decay_mult: 0
  1586. }
  1587. convolution_param {
  1588. num_output: 38
  1589. pad: 0
  1590. kernel_size: 1
  1591. weight_filler {
  1592. type: "gaussian"
  1593. std: 0.01
  1594. }
  1595. bias_filler {
  1596. type: "constant"
  1597. }
  1598. }
  1599. }
  1600. layer {
  1601. name: "Mconv7_stage3_L2"
  1602. type: "Convolution"
  1603. bottom: "Mconv6_stage3_L2"
  1604. top: "Mconv7_stage3_L2"
  1605. param {
  1606. lr_mult: 4.0
  1607. decay_mult: 1
  1608. }
  1609. param {
  1610. lr_mult: 8.0
  1611. decay_mult: 0
  1612. }
  1613. convolution_param {
  1614. num_output: 19
  1615. pad: 0
  1616. kernel_size: 1
  1617. weight_filler {
  1618. type: "gaussian"
  1619. std: 0.01
  1620. }
  1621. bias_filler {
  1622. type: "constant"
  1623. }
  1624. }
  1625. }
  1626. layer {
  1627. name: "concat_stage4"
  1628. type: "Concat"
  1629. bottom: "Mconv7_stage3_L1"
  1630. bottom: "Mconv7_stage3_L2"
  1631. bottom: "conv4_4_CPM"
  1632. top: "concat_stage4"
  1633. concat_param {
  1634. axis: 1
  1635. }
  1636. }
  1637. layer {
  1638. name: "Mconv1_stage4_L1"
  1639. type: "Convolution"
  1640. bottom: "concat_stage4"
  1641. top: "Mconv1_stage4_L1"
  1642. param {
  1643. lr_mult: 4.0
  1644. decay_mult: 1
  1645. }
  1646. param {
  1647. lr_mult: 8.0
  1648. decay_mult: 0
  1649. }
  1650. convolution_param {
  1651. num_output: 128
  1652. pad: 3
  1653. kernel_size: 7
  1654. weight_filler {
  1655. type: "gaussian"
  1656. std: 0.01
  1657. }
  1658. bias_filler {
  1659. type: "constant"
  1660. }
  1661. }
  1662. }
  1663. layer {
  1664. name: "Mrelu1_stage4_L1"
  1665. type: "ReLU"
  1666. bottom: "Mconv1_stage4_L1"
  1667. top: "Mconv1_stage4_L1"
  1668. }
  1669. layer {
  1670. name: "Mconv1_stage4_L2"
  1671. type: "Convolution"
  1672. bottom: "concat_stage4"
  1673. top: "Mconv1_stage4_L2"
  1674. param {
  1675. lr_mult: 4.0
  1676. decay_mult: 1
  1677. }
  1678. param {
  1679. lr_mult: 8.0
  1680. decay_mult: 0
  1681. }
  1682. convolution_param {
  1683. num_output: 128
  1684. pad: 3
  1685. kernel_size: 7
  1686. weight_filler {
  1687. type: "gaussian"
  1688. std: 0.01
  1689. }
  1690. bias_filler {
  1691. type: "constant"
  1692. }
  1693. }
  1694. }
  1695. layer {
  1696. name: "Mrelu1_stage4_L2"
  1697. type: "ReLU"
  1698. bottom: "Mconv1_stage4_L2"
  1699. top: "Mconv1_stage4_L2"
  1700. }
  1701. layer {
  1702. name: "Mconv2_stage4_L1"
  1703. type: "Convolution"
  1704. bottom: "Mconv1_stage4_L1"
  1705. top: "Mconv2_stage4_L1"
  1706. param {
  1707. lr_mult: 4.0
  1708. decay_mult: 1
  1709. }
  1710. param {
  1711. lr_mult: 8.0
  1712. decay_mult: 0
  1713. }
  1714. convolution_param {
  1715. num_output: 128
  1716. pad: 3
  1717. kernel_size: 7
  1718. weight_filler {
  1719. type: "gaussian"
  1720. std: 0.01
  1721. }
  1722. bias_filler {
  1723. type: "constant"
  1724. }
  1725. }
  1726. }
  1727. layer {
  1728. name: "Mrelu2_stage4_L1"
  1729. type: "ReLU"
  1730. bottom: "Mconv2_stage4_L1"
  1731. top: "Mconv2_stage4_L1"
  1732. }
  1733. layer {
  1734. name: "Mconv2_stage4_L2"
  1735. type: "Convolution"
  1736. bottom: "Mconv1_stage4_L2"
  1737. top: "Mconv2_stage4_L2"
  1738. param {
  1739. lr_mult: 4.0
  1740. decay_mult: 1
  1741. }
  1742. param {
  1743. lr_mult: 8.0
  1744. decay_mult: 0
  1745. }
  1746. convolution_param {
  1747. num_output: 128
  1748. pad: 3
  1749. kernel_size: 7
  1750. weight_filler {
  1751. type: "gaussian"
  1752. std: 0.01
  1753. }
  1754. bias_filler {
  1755. type: "constant"
  1756. }
  1757. }
  1758. }
  1759. layer {
  1760. name: "Mrelu2_stage4_L2"
  1761. type: "ReLU"
  1762. bottom: "Mconv2_stage4_L2"
  1763. top: "Mconv2_stage4_L2"
  1764. }
  1765. layer {
  1766. name: "Mconv3_stage4_L1"
  1767. type: "Convolution"
  1768. bottom: "Mconv2_stage4_L1"
  1769. top: "Mconv3_stage4_L1"
  1770. param {
  1771. lr_mult: 4.0
  1772. decay_mult: 1
  1773. }
  1774. param {
  1775. lr_mult: 8.0
  1776. decay_mult: 0
  1777. }
  1778. convolution_param {
  1779. num_output: 128
  1780. pad: 3
  1781. kernel_size: 7
  1782. weight_filler {
  1783. type: "gaussian"
  1784. std: 0.01
  1785. }
  1786. bias_filler {
  1787. type: "constant"
  1788. }
  1789. }
  1790. }
  1791. layer {
  1792. name: "Mrelu3_stage4_L1"
  1793. type: "ReLU"
  1794. bottom: "Mconv3_stage4_L1"
  1795. top: "Mconv3_stage4_L1"
  1796. }
  1797. layer {
  1798. name: "Mconv3_stage4_L2"
  1799. type: "Convolution"
  1800. bottom: "Mconv2_stage4_L2"
  1801. top: "Mconv3_stage4_L2"
  1802. param {
  1803. lr_mult: 4.0
  1804. decay_mult: 1
  1805. }
  1806. param {
  1807. lr_mult: 8.0
  1808. decay_mult: 0
  1809. }
  1810. convolution_param {
  1811. num_output: 128
  1812. pad: 3
  1813. kernel_size: 7
  1814. weight_filler {
  1815. type: "gaussian"
  1816. std: 0.01
  1817. }
  1818. bias_filler {
  1819. type: "constant"
  1820. }
  1821. }
  1822. }
  1823. layer {
  1824. name: "Mrelu3_stage4_L2"
  1825. type: "ReLU"
  1826. bottom: "Mconv3_stage4_L2"
  1827. top: "Mconv3_stage4_L2"
  1828. }
  1829. layer {
  1830. name: "Mconv4_stage4_L1"
  1831. type: "Convolution"
  1832. bottom: "Mconv3_stage4_L1"
  1833. top: "Mconv4_stage4_L1"
  1834. param {
  1835. lr_mult: 4.0
  1836. decay_mult: 1
  1837. }
  1838. param {
  1839. lr_mult: 8.0
  1840. decay_mult: 0
  1841. }
  1842. convolution_param {
  1843. num_output: 128
  1844. pad: 3
  1845. kernel_size: 7
  1846. weight_filler {
  1847. type: "gaussian"
  1848. std: 0.01
  1849. }
  1850. bias_filler {
  1851. type: "constant"
  1852. }
  1853. }
  1854. }
  1855. layer {
  1856. name: "Mrelu4_stage4_L1"
  1857. type: "ReLU"
  1858. bottom: "Mconv4_stage4_L1"
  1859. top: "Mconv4_stage4_L1"
  1860. }
  1861. layer {
  1862. name: "Mconv4_stage4_L2"
  1863. type: "Convolution"
  1864. bottom: "Mconv3_stage4_L2"
  1865. top: "Mconv4_stage4_L2"
  1866. param {
  1867. lr_mult: 4.0
  1868. decay_mult: 1
  1869. }
  1870. param {
  1871. lr_mult: 8.0
  1872. decay_mult: 0
  1873. }
  1874. convolution_param {
  1875. num_output: 128
  1876. pad: 3
  1877. kernel_size: 7
  1878. weight_filler {
  1879. type: "gaussian"
  1880. std: 0.01
  1881. }
  1882. bias_filler {
  1883. type: "constant"
  1884. }
  1885. }
  1886. }
  1887. layer {
  1888. name: "Mrelu4_stage4_L2"
  1889. type: "ReLU"
  1890. bottom: "Mconv4_stage4_L2"
  1891. top: "Mconv4_stage4_L2"
  1892. }
  1893. layer {
  1894. name: "Mconv5_stage4_L1"
  1895. type: "Convolution"
  1896. bottom: "Mconv4_stage4_L1"
  1897. top: "Mconv5_stage4_L1"
  1898. param {
  1899. lr_mult: 4.0
  1900. decay_mult: 1
  1901. }
  1902. param {
  1903. lr_mult: 8.0
  1904. decay_mult: 0
  1905. }
  1906. convolution_param {
  1907. num_output: 128
  1908. pad: 3
  1909. kernel_size: 7
  1910. weight_filler {
  1911. type: "gaussian"
  1912. std: 0.01
  1913. }
  1914. bias_filler {
  1915. type: "constant"
  1916. }
  1917. }
  1918. }
  1919. layer {
  1920. name: "Mrelu5_stage4_L1"
  1921. type: "ReLU"
  1922. bottom: "Mconv5_stage4_L1"
  1923. top: "Mconv5_stage4_L1"
  1924. }
  1925. layer {
  1926. name: "Mconv5_stage4_L2"
  1927. type: "Convolution"
  1928. bottom: "Mconv4_stage4_L2"
  1929. top: "Mconv5_stage4_L2"
  1930. param {
  1931. lr_mult: 4.0
  1932. decay_mult: 1
  1933. }
  1934. param {
  1935. lr_mult: 8.0
  1936. decay_mult: 0
  1937. }
  1938. convolution_param {
  1939. num_output: 128
  1940. pad: 3
  1941. kernel_size: 7
  1942. weight_filler {
  1943. type: "gaussian"
  1944. std: 0.01
  1945. }
  1946. bias_filler {
  1947. type: "constant"
  1948. }
  1949. }
  1950. }
  1951. layer {
  1952. name: "Mrelu5_stage4_L2"
  1953. type: "ReLU"
  1954. bottom: "Mconv5_stage4_L2"
  1955. top: "Mconv5_stage4_L2"
  1956. }
  1957. layer {
  1958. name: "Mconv6_stage4_L1"
  1959. type: "Convolution"
  1960. bottom: "Mconv5_stage4_L1"
  1961. top: "Mconv6_stage4_L1"
  1962. param {
  1963. lr_mult: 4.0
  1964. decay_mult: 1
  1965. }
  1966. param {
  1967. lr_mult: 8.0
  1968. decay_mult: 0
  1969. }
  1970. convolution_param {
  1971. num_output: 128
  1972. pad: 0
  1973. kernel_size: 1
  1974. weight_filler {
  1975. type: "gaussian"
  1976. std: 0.01
  1977. }
  1978. bias_filler {
  1979. type: "constant"
  1980. }
  1981. }
  1982. }
  1983. layer {
  1984. name: "Mrelu6_stage4_L1"
  1985. type: "ReLU"
  1986. bottom: "Mconv6_stage4_L1"
  1987. top: "Mconv6_stage4_L1"
  1988. }
  1989. layer {
  1990. name: "Mconv6_stage4_L2"
  1991. type: "Convolution"
  1992. bottom: "Mconv5_stage4_L2"
  1993. top: "Mconv6_stage4_L2"
  1994. param {
  1995. lr_mult: 4.0
  1996. decay_mult: 1
  1997. }
  1998. param {
  1999. lr_mult: 8.0
  2000. decay_mult: 0
  2001. }
  2002. convolution_param {
  2003. num_output: 128
  2004. pad: 0
  2005. kernel_size: 1
  2006. weight_filler {
  2007. type: "gaussian"
  2008. std: 0.01
  2009. }
  2010. bias_filler {
  2011. type: "constant"
  2012. }
  2013. }
  2014. }
  2015. layer {
  2016. name: "Mrelu6_stage4_L2"
  2017. type: "ReLU"
  2018. bottom: "Mconv6_stage4_L2"
  2019. top: "Mconv6_stage4_L2"
  2020. }
  2021. layer {
  2022. name: "Mconv7_stage4_L1"
  2023. type: "Convolution"
  2024. bottom: "Mconv6_stage4_L1"
  2025. top: "Mconv7_stage4_L1"
  2026. param {
  2027. lr_mult: 4.0
  2028. decay_mult: 1
  2029. }
  2030. param {
  2031. lr_mult: 8.0
  2032. decay_mult: 0
  2033. }
  2034. convolution_param {
  2035. num_output: 38
  2036. pad: 0
  2037. kernel_size: 1
  2038. weight_filler {
  2039. type: "gaussian"
  2040. std: 0.01
  2041. }
  2042. bias_filler {
  2043. type: "constant"
  2044. }
  2045. }
  2046. }
  2047. layer {
  2048. name: "Mconv7_stage4_L2"
  2049. type: "Convolution"
  2050. bottom: "Mconv6_stage4_L2"
  2051. top: "Mconv7_stage4_L2"
  2052. param {
  2053. lr_mult: 4.0
  2054. decay_mult: 1
  2055. }
  2056. param {
  2057. lr_mult: 8.0
  2058. decay_mult: 0
  2059. }
  2060. convolution_param {
  2061. num_output: 19
  2062. pad: 0
  2063. kernel_size: 1
  2064. weight_filler {
  2065. type: "gaussian"
  2066. std: 0.01
  2067. }
  2068. bias_filler {
  2069. type: "constant"
  2070. }
  2071. }
  2072. }
  2073. layer {
  2074. name: "concat_stage5"
  2075. type: "Concat"
  2076. bottom: "Mconv7_stage4_L1"
  2077. bottom: "Mconv7_stage4_L2"
  2078. bottom: "conv4_4_CPM"
  2079. top: "concat_stage5"
  2080. concat_param {
  2081. axis: 1
  2082. }
  2083. }
  2084. layer {
  2085. name: "Mconv1_stage5_L1"
  2086. type: "Convolution"
  2087. bottom: "concat_stage5"
  2088. top: "Mconv1_stage5_L1"
  2089. param {
  2090. lr_mult: 4.0
  2091. decay_mult: 1
  2092. }
  2093. param {
  2094. lr_mult: 8.0
  2095. decay_mult: 0
  2096. }
  2097. convolution_param {
  2098. num_output: 128
  2099. pad: 3
  2100. kernel_size: 7
  2101. weight_filler {
  2102. type: "gaussian"
  2103. std: 0.01
  2104. }
  2105. bias_filler {
  2106. type: "constant"
  2107. }
  2108. }
  2109. }
  2110. layer {
  2111. name: "Mrelu1_stage5_L1"
  2112. type: "ReLU"
  2113. bottom: "Mconv1_stage5_L1"
  2114. top: "Mconv1_stage5_L1"
  2115. }
  2116. layer {
  2117. name: "Mconv1_stage5_L2"
  2118. type: "Convolution"
  2119. bottom: "concat_stage5"
  2120. top: "Mconv1_stage5_L2"
  2121. param {
  2122. lr_mult: 4.0
  2123. decay_mult: 1
  2124. }
  2125. param {
  2126. lr_mult: 8.0
  2127. decay_mult: 0
  2128. }
  2129. convolution_param {
  2130. num_output: 128
  2131. pad: 3
  2132. kernel_size: 7
  2133. weight_filler {
  2134. type: "gaussian"
  2135. std: 0.01
  2136. }
  2137. bias_filler {
  2138. type: "constant"
  2139. }
  2140. }
  2141. }
  2142. layer {
  2143. name: "Mrelu1_stage5_L2"
  2144. type: "ReLU"
  2145. bottom: "Mconv1_stage5_L2"
  2146. top: "Mconv1_stage5_L2"
  2147. }
  2148. layer {
  2149. name: "Mconv2_stage5_L1"
  2150. type: "Convolution"
  2151. bottom: "Mconv1_stage5_L1"
  2152. top: "Mconv2_stage5_L1"
  2153. param {
  2154. lr_mult: 4.0
  2155. decay_mult: 1
  2156. }
  2157. param {
  2158. lr_mult: 8.0
  2159. decay_mult: 0
  2160. }
  2161. convolution_param {
  2162. num_output: 128
  2163. pad: 3
  2164. kernel_size: 7
  2165. weight_filler {
  2166. type: "gaussian"
  2167. std: 0.01
  2168. }
  2169. bias_filler {
  2170. type: "constant"
  2171. }
  2172. }
  2173. }
  2174. layer {
  2175. name: "Mrelu2_stage5_L1"
  2176. type: "ReLU"
  2177. bottom: "Mconv2_stage5_L1"
  2178. top: "Mconv2_stage5_L1"
  2179. }
  2180. layer {
  2181. name: "Mconv2_stage5_L2"
  2182. type: "Convolution"
  2183. bottom: "Mconv1_stage5_L2"
  2184. top: "Mconv2_stage5_L2"
  2185. param {
  2186. lr_mult: 4.0
  2187. decay_mult: 1
  2188. }
  2189. param {
  2190. lr_mult: 8.0
  2191. decay_mult: 0
  2192. }
  2193. convolution_param {
  2194. num_output: 128
  2195. pad: 3
  2196. kernel_size: 7
  2197. weight_filler {
  2198. type: "gaussian"
  2199. std: 0.01
  2200. }
  2201. bias_filler {
  2202. type: "constant"
  2203. }
  2204. }
  2205. }
  2206. layer {
  2207. name: "Mrelu2_stage5_L2"
  2208. type: "ReLU"
  2209. bottom: "Mconv2_stage5_L2"
  2210. top: "Mconv2_stage5_L2"
  2211. }
  2212. layer {
  2213. name: "Mconv3_stage5_L1"
  2214. type: "Convolution"
  2215. bottom: "Mconv2_stage5_L1"
  2216. top: "Mconv3_stage5_L1"
  2217. param {
  2218. lr_mult: 4.0
  2219. decay_mult: 1
  2220. }
  2221. param {
  2222. lr_mult: 8.0
  2223. decay_mult: 0
  2224. }
  2225. convolution_param {
  2226. num_output: 128
  2227. pad: 3
  2228. kernel_size: 7
  2229. weight_filler {
  2230. type: "gaussian"
  2231. std: 0.01
  2232. }
  2233. bias_filler {
  2234. type: "constant"
  2235. }
  2236. }
  2237. }
  2238. layer {
  2239. name: "Mrelu3_stage5_L1"
  2240. type: "ReLU"
  2241. bottom: "Mconv3_stage5_L1"
  2242. top: "Mconv3_stage5_L1"
  2243. }
  2244. layer {
  2245. name: "Mconv3_stage5_L2"
  2246. type: "Convolution"
  2247. bottom: "Mconv2_stage5_L2"
  2248. top: "Mconv3_stage5_L2"
  2249. param {
  2250. lr_mult: 4.0
  2251. decay_mult: 1
  2252. }
  2253. param {
  2254. lr_mult: 8.0
  2255. decay_mult: 0
  2256. }
  2257. convolution_param {
  2258. num_output: 128
  2259. pad: 3
  2260. kernel_size: 7
  2261. weight_filler {
  2262. type: "gaussian"
  2263. std: 0.01
  2264. }
  2265. bias_filler {
  2266. type: "constant"
  2267. }
  2268. }
  2269. }
  2270. layer {
  2271. name: "Mrelu3_stage5_L2"
  2272. type: "ReLU"
  2273. bottom: "Mconv3_stage5_L2"
  2274. top: "Mconv3_stage5_L2"
  2275. }
  2276. layer {
  2277. name: "Mconv4_stage5_L1"
  2278. type: "Convolution"
  2279. bottom: "Mconv3_stage5_L1"
  2280. top: "Mconv4_stage5_L1"
  2281. param {
  2282. lr_mult: 4.0
  2283. decay_mult: 1
  2284. }
  2285. param {
  2286. lr_mult: 8.0
  2287. decay_mult: 0
  2288. }
  2289. convolution_param {
  2290. num_output: 128
  2291. pad: 3
  2292. kernel_size: 7
  2293. weight_filler {
  2294. type: "gaussian"
  2295. std: 0.01
  2296. }
  2297. bias_filler {
  2298. type: "constant"
  2299. }
  2300. }
  2301. }
  2302. layer {
  2303. name: "Mrelu4_stage5_L1"
  2304. type: "ReLU"
  2305. bottom: "Mconv4_stage5_L1"
  2306. top: "Mconv4_stage5_L1"
  2307. }
  2308. layer {
  2309. name: "Mconv4_stage5_L2"
  2310. type: "Convolution"
  2311. bottom: "Mconv3_stage5_L2"
  2312. top: "Mconv4_stage5_L2"
  2313. param {
  2314. lr_mult: 4.0
  2315. decay_mult: 1
  2316. }
  2317. param {
  2318. lr_mult: 8.0
  2319. decay_mult: 0
  2320. }
  2321. convolution_param {
  2322. num_output: 128
  2323. pad: 3
  2324. kernel_size: 7
  2325. weight_filler {
  2326. type: "gaussian"
  2327. std: 0.01
  2328. }
  2329. bias_filler {
  2330. type: "constant"
  2331. }
  2332. }
  2333. }
  2334. layer {
  2335. name: "Mrelu4_stage5_L2"
  2336. type: "ReLU"
  2337. bottom: "Mconv4_stage5_L2"
  2338. top: "Mconv4_stage5_L2"
  2339. }
  2340. layer {
  2341. name: "Mconv5_stage5_L1"
  2342. type: "Convolution"
  2343. bottom: "Mconv4_stage5_L1"
  2344. top: "Mconv5_stage5_L1"
  2345. param {
  2346. lr_mult: 4.0
  2347. decay_mult: 1
  2348. }
  2349. param {
  2350. lr_mult: 8.0
  2351. decay_mult: 0
  2352. }
  2353. convolution_param {
  2354. num_output: 128
  2355. pad: 3
  2356. kernel_size: 7
  2357. weight_filler {
  2358. type: "gaussian"
  2359. std: 0.01
  2360. }
  2361. bias_filler {
  2362. type: "constant"
  2363. }
  2364. }
  2365. }
  2366. layer {
  2367. name: "Mrelu5_stage5_L1"
  2368. type: "ReLU"
  2369. bottom: "Mconv5_stage5_L1"
  2370. top: "Mconv5_stage5_L1"
  2371. }
  2372. layer {
  2373. name: "Mconv5_stage5_L2"
  2374. type: "Convolution"
  2375. bottom: "Mconv4_stage5_L2"
  2376. top: "Mconv5_stage5_L2"
  2377. param {
  2378. lr_mult: 4.0
  2379. decay_mult: 1
  2380. }
  2381. param {
  2382. lr_mult: 8.0
  2383. decay_mult: 0
  2384. }
  2385. convolution_param {
  2386. num_output: 128
  2387. pad: 3
  2388. kernel_size: 7
  2389. weight_filler {
  2390. type: "gaussian"
  2391. std: 0.01
  2392. }
  2393. bias_filler {
  2394. type: "constant"
  2395. }
  2396. }
  2397. }
  2398. layer {
  2399. name: "Mrelu5_stage5_L2"
  2400. type: "ReLU"
  2401. bottom: "Mconv5_stage5_L2"
  2402. top: "Mconv5_stage5_L2"
  2403. }
  2404. layer {
  2405. name: "Mconv6_stage5_L1"
  2406. type: "Convolution"
  2407. bottom: "Mconv5_stage5_L1"
  2408. top: "Mconv6_stage5_L1"
  2409. param {
  2410. lr_mult: 4.0
  2411. decay_mult: 1
  2412. }
  2413. param {
  2414. lr_mult: 8.0
  2415. decay_mult: 0
  2416. }
  2417. convolution_param {
  2418. num_output: 128
  2419. pad: 0
  2420. kernel_size: 1
  2421. weight_filler {
  2422. type: "gaussian"
  2423. std: 0.01
  2424. }
  2425. bias_filler {
  2426. type: "constant"
  2427. }
  2428. }
  2429. }
  2430. layer {
  2431. name: "Mrelu6_stage5_L1"
  2432. type: "ReLU"
  2433. bottom: "Mconv6_stage5_L1"
  2434. top: "Mconv6_stage5_L1"
  2435. }
  2436. layer {
  2437. name: "Mconv6_stage5_L2"
  2438. type: "Convolution"
  2439. bottom: "Mconv5_stage5_L2"
  2440. top: "Mconv6_stage5_L2"
  2441. param {
  2442. lr_mult: 4.0
  2443. decay_mult: 1
  2444. }
  2445. param {
  2446. lr_mult: 8.0
  2447. decay_mult: 0
  2448. }
  2449. convolution_param {
  2450. num_output: 128
  2451. pad: 0
  2452. kernel_size: 1
  2453. weight_filler {
  2454. type: "gaussian"
  2455. std: 0.01
  2456. }
  2457. bias_filler {
  2458. type: "constant"
  2459. }
  2460. }
  2461. }
  2462. layer {
  2463. name: "Mrelu6_stage5_L2"
  2464. type: "ReLU"
  2465. bottom: "Mconv6_stage5_L2"
  2466. top: "Mconv6_stage5_L2"
  2467. }
  2468. layer {
  2469. name: "Mconv7_stage5_L1"
  2470. type: "Convolution"
  2471. bottom: "Mconv6_stage5_L1"
  2472. top: "Mconv7_stage5_L1"
  2473. param {
  2474. lr_mult: 4.0
  2475. decay_mult: 1
  2476. }
  2477. param {
  2478. lr_mult: 8.0
  2479. decay_mult: 0
  2480. }
  2481. convolution_param {
  2482. num_output: 38
  2483. pad: 0
  2484. kernel_size: 1
  2485. weight_filler {
  2486. type: "gaussian"
  2487. std: 0.01
  2488. }
  2489. bias_filler {
  2490. type: "constant"
  2491. }
  2492. }
  2493. }
  2494. layer {
  2495. name: "Mconv7_stage5_L2"
  2496. type: "Convolution"
  2497. bottom: "Mconv6_stage5_L2"
  2498. top: "Mconv7_stage5_L2"
  2499. param {
  2500. lr_mult: 4.0
  2501. decay_mult: 1
  2502. }
  2503. param {
  2504. lr_mult: 8.0
  2505. decay_mult: 0
  2506. }
  2507. convolution_param {
  2508. num_output: 19
  2509. pad: 0
  2510. kernel_size: 1
  2511. weight_filler {
  2512. type: "gaussian"
  2513. std: 0.01
  2514. }
  2515. bias_filler {
  2516. type: "constant"
  2517. }
  2518. }
  2519. }
  2520. layer {
  2521. name: "concat_stage6"
  2522. type: "Concat"
  2523. bottom: "Mconv7_stage5_L1"
  2524. bottom: "Mconv7_stage5_L2"
  2525. bottom: "conv4_4_CPM"
  2526. top: "concat_stage6"
  2527. concat_param {
  2528. axis: 1
  2529. }
  2530. }
  2531. layer {
  2532. name: "Mconv1_stage6_L1"
  2533. type: "Convolution"
  2534. bottom: "concat_stage6"
  2535. top: "Mconv1_stage6_L1"
  2536. param {
  2537. lr_mult: 4.0
  2538. decay_mult: 1
  2539. }
  2540. param {
  2541. lr_mult: 8.0
  2542. decay_mult: 0
  2543. }
  2544. convolution_param {
  2545. num_output: 128
  2546. pad: 3
  2547. kernel_size: 7
  2548. weight_filler {
  2549. type: "gaussian"
  2550. std: 0.01
  2551. }
  2552. bias_filler {
  2553. type: "constant"
  2554. }
  2555. }
  2556. }
  2557. layer {
  2558. name: "Mrelu1_stage6_L1"
  2559. type: "ReLU"
  2560. bottom: "Mconv1_stage6_L1"
  2561. top: "Mconv1_stage6_L1"
  2562. }
  2563. layer {
  2564. name: "Mconv1_stage6_L2"
  2565. type: "Convolution"
  2566. bottom: "concat_stage6"
  2567. top: "Mconv1_stage6_L2"
  2568. param {
  2569. lr_mult: 4.0
  2570. decay_mult: 1
  2571. }
  2572. param {
  2573. lr_mult: 8.0
  2574. decay_mult: 0
  2575. }
  2576. convolution_param {
  2577. num_output: 128
  2578. pad: 3
  2579. kernel_size: 7
  2580. weight_filler {
  2581. type: "gaussian"
  2582. std: 0.01
  2583. }
  2584. bias_filler {
  2585. type: "constant"
  2586. }
  2587. }
  2588. }
  2589. layer {
  2590. name: "Mrelu1_stage6_L2"
  2591. type: "ReLU"
  2592. bottom: "Mconv1_stage6_L2"
  2593. top: "Mconv1_stage6_L2"
  2594. }
  2595. layer {
  2596. name: "Mconv2_stage6_L1"
  2597. type: "Convolution"
  2598. bottom: "Mconv1_stage6_L1"
  2599. top: "Mconv2_stage6_L1"
  2600. param {
  2601. lr_mult: 4.0
  2602. decay_mult: 1
  2603. }
  2604. param {
  2605. lr_mult: 8.0
  2606. decay_mult: 0
  2607. }
  2608. convolution_param {
  2609. num_output: 128
  2610. pad: 3
  2611. kernel_size: 7
  2612. weight_filler {
  2613. type: "gaussian"
  2614. std: 0.01
  2615. }
  2616. bias_filler {
  2617. type: "constant"
  2618. }
  2619. }
  2620. }
  2621. layer {
  2622. name: "Mrelu2_stage6_L1"
  2623. type: "ReLU"
  2624. bottom: "Mconv2_stage6_L1"
  2625. top: "Mconv2_stage6_L1"
  2626. }
  2627. layer {
  2628. name: "Mconv2_stage6_L2"
  2629. type: "Convolution"
  2630. bottom: "Mconv1_stage6_L2"
  2631. top: "Mconv2_stage6_L2"
  2632. param {
  2633. lr_mult: 4.0
  2634. decay_mult: 1
  2635. }
  2636. param {
  2637. lr_mult: 8.0
  2638. decay_mult: 0
  2639. }
  2640. convolution_param {
  2641. num_output: 128
  2642. pad: 3
  2643. kernel_size: 7
  2644. weight_filler {
  2645. type: "gaussian"
  2646. std: 0.01
  2647. }
  2648. bias_filler {
  2649. type: "constant"
  2650. }
  2651. }
  2652. }
  2653. layer {
  2654. name: "Mrelu2_stage6_L2"
  2655. type: "ReLU"
  2656. bottom: "Mconv2_stage6_L2"
  2657. top: "Mconv2_stage6_L2"
  2658. }
  2659. layer {
  2660. name: "Mconv3_stage6_L1"
  2661. type: "Convolution"
  2662. bottom: "Mconv2_stage6_L1"
  2663. top: "Mconv3_stage6_L1"
  2664. param {
  2665. lr_mult: 4.0
  2666. decay_mult: 1
  2667. }
  2668. param {
  2669. lr_mult: 8.0
  2670. decay_mult: 0
  2671. }
  2672. convolution_param {
  2673. num_output: 128
  2674. pad: 3
  2675. kernel_size: 7
  2676. weight_filler {
  2677. type: "gaussian"
  2678. std: 0.01
  2679. }
  2680. bias_filler {
  2681. type: "constant"
  2682. }
  2683. }
  2684. }
  2685. layer {
  2686. name: "Mrelu3_stage6_L1"
  2687. type: "ReLU"
  2688. bottom: "Mconv3_stage6_L1"
  2689. top: "Mconv3_stage6_L1"
  2690. }
  2691. layer {
  2692. name: "Mconv3_stage6_L2"
  2693. type: "Convolution"
  2694. bottom: "Mconv2_stage6_L2"
  2695. top: "Mconv3_stage6_L2"
  2696. param {
  2697. lr_mult: 4.0
  2698. decay_mult: 1
  2699. }
  2700. param {
  2701. lr_mult: 8.0
  2702. decay_mult: 0
  2703. }
  2704. convolution_param {
  2705. num_output: 128
  2706. pad: 3
  2707. kernel_size: 7
  2708. weight_filler {
  2709. type: "gaussian"
  2710. std: 0.01
  2711. }
  2712. bias_filler {
  2713. type: "constant"
  2714. }
  2715. }
  2716. }
  2717. layer {
  2718. name: "Mrelu3_stage6_L2"
  2719. type: "ReLU"
  2720. bottom: "Mconv3_stage6_L2"
  2721. top: "Mconv3_stage6_L2"
  2722. }
  2723. layer {
  2724. name: "Mconv4_stage6_L1"
  2725. type: "Convolution"
  2726. bottom: "Mconv3_stage6_L1"
  2727. top: "Mconv4_stage6_L1"
  2728. param {
  2729. lr_mult: 4.0
  2730. decay_mult: 1
  2731. }
  2732. param {
  2733. lr_mult: 8.0
  2734. decay_mult: 0
  2735. }
  2736. convolution_param {
  2737. num_output: 128
  2738. pad: 3
  2739. kernel_size: 7
  2740. weight_filler {
  2741. type: "gaussian"
  2742. std: 0.01
  2743. }
  2744. bias_filler {
  2745. type: "constant"
  2746. }
  2747. }
  2748. }
  2749. layer {
  2750. name: "Mrelu4_stage6_L1"
  2751. type: "ReLU"
  2752. bottom: "Mconv4_stage6_L1"
  2753. top: "Mconv4_stage6_L1"
  2754. }
  2755. layer {
  2756. name: "Mconv4_stage6_L2"
  2757. type: "Convolution"
  2758. bottom: "Mconv3_stage6_L2"
  2759. top: "Mconv4_stage6_L2"
  2760. param {
  2761. lr_mult: 4.0
  2762. decay_mult: 1
  2763. }
  2764. param {
  2765. lr_mult: 8.0
  2766. decay_mult: 0
  2767. }
  2768. convolution_param {
  2769. num_output: 128
  2770. pad: 3
  2771. kernel_size: 7
  2772. weight_filler {
  2773. type: "gaussian"
  2774. std: 0.01
  2775. }
  2776. bias_filler {
  2777. type: "constant"
  2778. }
  2779. }
  2780. }
  2781. layer {
  2782. name: "Mrelu4_stage6_L2"
  2783. type: "ReLU"
  2784. bottom: "Mconv4_stage6_L2"
  2785. top: "Mconv4_stage6_L2"
  2786. }
  2787. layer {
  2788. name: "Mconv5_stage6_L1"
  2789. type: "Convolution"
  2790. bottom: "Mconv4_stage6_L1"
  2791. top: "Mconv5_stage6_L1"
  2792. param {
  2793. lr_mult: 4.0
  2794. decay_mult: 1
  2795. }
  2796. param {
  2797. lr_mult: 8.0
  2798. decay_mult: 0
  2799. }
  2800. convolution_param {
  2801. num_output: 128
  2802. pad: 3
  2803. kernel_size: 7
  2804. weight_filler {
  2805. type: "gaussian"
  2806. std: 0.01
  2807. }
  2808. bias_filler {
  2809. type: "constant"
  2810. }
  2811. }
  2812. }
  2813. layer {
  2814. name: "Mrelu5_stage6_L1"
  2815. type: "ReLU"
  2816. bottom: "Mconv5_stage6_L1"
  2817. top: "Mconv5_stage6_L1"
  2818. }
  2819. layer {
  2820. name: "Mconv5_stage6_L2"
  2821. type: "Convolution"
  2822. bottom: "Mconv4_stage6_L2"
  2823. top: "Mconv5_stage6_L2"
  2824. param {
  2825. lr_mult: 4.0
  2826. decay_mult: 1
  2827. }
  2828. param {
  2829. lr_mult: 8.0
  2830. decay_mult: 0
  2831. }
  2832. convolution_param {
  2833. num_output: 128
  2834. pad: 3
  2835. kernel_size: 7
  2836. weight_filler {
  2837. type: "gaussian"
  2838. std: 0.01
  2839. }
  2840. bias_filler {
  2841. type: "constant"
  2842. }
  2843. }
  2844. }
  2845. layer {
  2846. name: "Mrelu5_stage6_L2"
  2847. type: "ReLU"
  2848. bottom: "Mconv5_stage6_L2"
  2849. top: "Mconv5_stage6_L2"
  2850. }
  2851. layer {
  2852. name: "Mconv6_stage6_L1"
  2853. type: "Convolution"
  2854. bottom: "Mconv5_stage6_L1"
  2855. top: "Mconv6_stage6_L1"
  2856. param {
  2857. lr_mult: 4.0
  2858. decay_mult: 1
  2859. }
  2860. param {
  2861. lr_mult: 8.0
  2862. decay_mult: 0
  2863. }
  2864. convolution_param {
  2865. num_output: 128
  2866. pad: 0
  2867. kernel_size: 1
  2868. weight_filler {
  2869. type: "gaussian"
  2870. std: 0.01
  2871. }
  2872. bias_filler {
  2873. type: "constant"
  2874. }
  2875. }
  2876. }
  2877. layer {
  2878. name: "Mrelu6_stage6_L1"
  2879. type: "ReLU"
  2880. bottom: "Mconv6_stage6_L1"
  2881. top: "Mconv6_stage6_L1"
  2882. }
  2883. layer {
  2884. name: "Mconv6_stage6_L2"
  2885. type: "Convolution"
  2886. bottom: "Mconv5_stage6_L2"
  2887. top: "Mconv6_stage6_L2"
  2888. param {
  2889. lr_mult: 4.0
  2890. decay_mult: 1
  2891. }
  2892. param {
  2893. lr_mult: 8.0
  2894. decay_mult: 0
  2895. }
  2896. convolution_param {
  2897. num_output: 128
  2898. pad: 0
  2899. kernel_size: 1
  2900. weight_filler {
  2901. type: "gaussian"
  2902. std: 0.01
  2903. }
  2904. bias_filler {
  2905. type: "constant"
  2906. }
  2907. }
  2908. }
  2909. layer {
  2910. name: "Mrelu6_stage6_L2"
  2911. type: "ReLU"
  2912. bottom: "Mconv6_stage6_L2"
  2913. top: "Mconv6_stage6_L2"
  2914. }
  2915. layer {
  2916. name: "Mconv7_stage6_L1"
  2917. type: "Convolution"
  2918. bottom: "Mconv6_stage6_L1"
  2919. top: "Mconv7_stage6_L1"
  2920. param {
  2921. lr_mult: 4.0
  2922. decay_mult: 1
  2923. }
  2924. param {
  2925. lr_mult: 8.0
  2926. decay_mult: 0
  2927. }
  2928. convolution_param {
  2929. num_output: 38
  2930. pad: 0
  2931. kernel_size: 1
  2932. weight_filler {
  2933. type: "gaussian"
  2934. std: 0.01
  2935. }
  2936. bias_filler {
  2937. type: "constant"
  2938. }
  2939. }
  2940. }
  2941. layer {
  2942. name: "Mconv7_stage6_L2"
  2943. type: "Convolution"
  2944. bottom: "Mconv6_stage6_L2"
  2945. top: "Mconv7_stage6_L2"
  2946. param {
  2947. lr_mult: 4.0
  2948. decay_mult: 1
  2949. }
  2950. param {
  2951. lr_mult: 8.0
  2952. decay_mult: 0
  2953. }
  2954. convolution_param {
  2955. num_output: 19
  2956. pad: 0
  2957. kernel_size: 1
  2958. weight_filler {
  2959. type: "gaussian"
  2960. std: 0.01
  2961. }
  2962. bias_filler {
  2963. type: "constant"
  2964. }
  2965. }
  2966. }
  2967. layer {
  2968. name: "concat_stage7"
  2969. type: "Concat"
  2970. bottom: "Mconv7_stage6_L2"
  2971. bottom: "Mconv7_stage6_L1"
  2972. # top: "concat_stage7"
  2973. top: "net_output"
  2974. concat_param {
  2975. axis: 1
  2976. }
  2977. }
Advertisement
Add Comment
Please, Sign In to add comment
Advertisement
Public Pastes
Advertisement
We use cookies for various purposes including analytics. By continuing to use Pastebin, you agree to our use of cookies as described in the Cookies Policy.  OK, I Understand
Not a member of Pastebin yet?
Sign Up, it unlocks many cool features!