API tools faq
paste
Guest User

Untitled

a guest
Jan 16th, 2018
96
0
Never
Add comment
Not a member of Pastebin yet? Sign Up, it unlocks many cool features!
text 15.64 KB | None | 0 0
raw download clone embed print report
  1. name: "Face-ResNet"
  2. layer {
  3. name: "conv1a"
  4. type: "Convolution"
  5. bottom: "data"
  6. top: "conv1a"
  7. param {
  8. lr_mult: 1
  9. decay_mult: 1
  10. }
  11. param {
  12. lr_mult: 2
  13. decay_mult: 0
  14. }
  15. convolution_param {
  16. num_output: 32
  17. kernel_size: 3
  18. stride: 1
  19. weight_filler {
  20. type: "xavier"
  21. }
  22. bias_filler {
  23. type: "constant"
  24. value: 0
  25. }
  26. }
  27. }
  28. layer {
  29. name: "relu1a"
  30. type: "PReLU"
  31. bottom: "conv1a"
  32. top: "conv1a"
  33. }
  34. layer {
  35. name: "conv1b"
  36. type: "Convolution"
  37. bottom: "conv1a"
  38. top: "conv1b"
  39. param {
  40. lr_mult: 1
  41. decay_mult: 1
  42. }
  43. param {
  44. lr_mult: 2
  45. decay_mult: 0
  46. }
  47. convolution_param {
  48. num_output: 64
  49. kernel_size: 3
  50. stride: 1
  51. weight_filler {
  52. type: "xavier"
  53. }
  54. bias_filler {
  55. type: "constant"
  56. value: 0
  57. }
  58. }
  59. }
  60. layer {
  61. name: "relu1b"
  62. type: "PReLU"
  63. bottom: "conv1b"
  64. top: "conv1b"
  65. }
  66. layer {
  67. name: "pool1b"
  68. type: "Pooling"
  69. bottom: "conv1b"
  70. top: "pool1b"
  71. pooling_param {
  72. pool: MAX
  73. kernel_size: 2
  74. stride: 2
  75. }
  76. }
  77. layer {
  78. name: "conv2_1"
  79. type: "Convolution"
  80. bottom: "pool1b"
  81. top: "conv2_1"
  82. param {
  83. lr_mult: 1
  84. decay_mult: 1
  85. }
  86. param {
  87. lr_mult: 0
  88. decay_mult: 0
  89. }
  90. convolution_param {
  91. num_output: 64
  92. kernel_size: 3
  93. stride: 1
  94. pad: 1
  95. weight_filler {
  96. type: "gaussian"
  97. std: 0.01
  98. }
  99. bias_filler {
  100. type: "constant"
  101. value: 0
  102. }
  103. }
  104. }
  105. layer {
  106. name: "relu2_1"
  107. type: "PReLU"
  108. bottom: "conv2_1"
  109. top: "conv2_1"
  110. }
  111. layer {
  112. name: "conv2_2"
  113. type: "Convolution"
  114. bottom: "conv2_1"
  115. top: "conv2_2"
  116. param {
  117. lr_mult: 1
  118. decay_mult: 1
  119. }
  120. param {
  121. lr_mult: 0
  122. decay_mult: 0
  123. }
  124. convolution_param {
  125. num_output: 64
  126. kernel_size: 3
  127. stride: 1
  128. pad: 1
  129. weight_filler {
  130. type: "gaussian"
  131. std: 0.01
  132. }
  133. bias_filler {
  134. type: "constant"
  135. value: 0
  136. }
  137. }
  138. }
  139. layer {
  140. name: "relu2_2"
  141. type: "PReLU"
  142. bottom: "conv2_2"
  143. top: "conv2_2"
  144. }
  145. layer {
  146. name: "res2_2"
  147. type: "Eltwise"
  148. bottom: "pool1b"
  149. bottom: "conv2_2"
  150. top: "res2_2"
  151. eltwise_param {
  152. operation: 1
  153. }
  154. }
  155. layer {
  156. name: "conv2"
  157. type: "Convolution"
  158. bottom: "res2_2"
  159. top: "conv2"
  160. param {
  161. lr_mult: 1
  162. decay_mult: 1
  163. }
  164. param {
  165. lr_mult: 2
  166. decay_mult: 0
  167. }
  168. convolution_param {
  169. num_output: 128
  170. kernel_size: 3
  171. stride: 1
  172. weight_filler {
  173. type: "xavier"
  174. }
  175. bias_filler {
  176. type: "constant"
  177. value: 0
  178. }
  179. }
  180. }
  181. layer {
  182. name: "relu2"
  183. type: "PReLU"
  184. bottom: "conv2"
  185. top: "conv2"
  186. }
  187. layer {
  188. name: "pool2"
  189. type: "Pooling"
  190. bottom: "conv2"
  191. top: "pool2"
  192. pooling_param {
  193. pool: MAX
  194. kernel_size: 2
  195. stride: 2
  196. }
  197. }
  198. layer {
  199. name: "conv3_1"
  200. type: "Convolution"
  201. bottom: "pool2"
  202. top: "conv3_1"
  203. param {
  204. lr_mult: 1
  205. decay_mult: 1
  206. }
  207. param {
  208. lr_mult: 0
  209. decay_mult: 0
  210. }
  211. convolution_param {
  212. num_output: 128
  213. kernel_size: 3
  214. stride: 1
  215. pad: 1
  216. weight_filler {
  217. type: "gaussian"
  218. std: 0.01
  219. }
  220. bias_filler {
  221. type: "constant"
  222. value: 0
  223. }
  224. }
  225. }
  226. layer {
  227. name: "relu3_1"
  228. type: "PReLU"
  229. bottom: "conv3_1"
  230. top: "conv3_1"
  231. }
  232. layer {
  233. name: "conv3_2"
  234. type: "Convolution"
  235. bottom: "conv3_1"
  236. top: "conv3_2"
  237. param {
  238. lr_mult: 1
  239. decay_mult: 1
  240. }
  241. param {
  242. lr_mult: 0
  243. decay_mult: 0
  244. }
  245. convolution_param {
  246. num_output: 128
  247. kernel_size: 3
  248. stride: 1
  249. pad: 1
  250. weight_filler {
  251. type: "gaussian"
  252. std: 0.01
  253. }
  254. bias_filler {
  255. type: "constant"
  256. value: 0
  257. }
  258. }
  259. }
  260. layer {
  261. name: "relu3_2"
  262. type: "PReLU"
  263. bottom: "conv3_2"
  264. top: "conv3_2"
  265. }
  266. layer {
  267. name: "res3_2"
  268. type: "Eltwise"
  269. bottom: "pool2"
  270. bottom: "conv3_2"
  271. top: "res3_2"
  272. eltwise_param {
  273. operation: 1
  274. }
  275. }
  276. layer {
  277. name: "conv3_3"
  278. type: "Convolution"
  279. bottom: "res3_2"
  280. top: "conv3_3"
  281. param {
  282. lr_mult: 1
  283. decay_mult: 1
  284. }
  285. param {
  286. lr_mult: 0
  287. decay_mult: 0
  288. }
  289. convolution_param {
  290. num_output: 128
  291. kernel_size: 3
  292. stride: 1
  293. pad: 1
  294. weight_filler {
  295. type: "gaussian"
  296. std: 0.01
  297. }
  298. bias_filler {
  299. type: "constant"
  300. value: 0
  301. }
  302. }
  303. }
  304. layer {
  305. name: "relu3_3"
  306. type: "PReLU"
  307. bottom: "conv3_3"
  308. top: "conv3_3"
  309. }
  310. layer {
  311. name: "conv3_4"
  312. type: "Convolution"
  313. bottom: "conv3_3"
  314. top: "conv3_4"
  315. param {
  316. lr_mult: 1
  317. decay_mult: 1
  318. }
  319. param {
  320. lr_mult: 0
  321. decay_mult: 0
  322. }
  323. convolution_param {
  324. num_output: 128
  325. kernel_size: 3
  326. stride: 1
  327. pad: 1
  328. weight_filler {
  329. type: "gaussian"
  330. std: 0.01
  331. }
  332. bias_filler {
  333. type: "constant"
  334. value: 0
  335. }
  336. }
  337. }
  338. layer {
  339. name: "relu3_4"
  340. type: "PReLU"
  341. bottom: "conv3_4"
  342. top: "conv3_4"
  343. }
  344. layer {
  345. name: "res3_4"
  346. type: "Eltwise"
  347. bottom: "res3_2"
  348. bottom: "conv3_4"
  349. top: "res3_4"
  350. eltwise_param {
  351. operation: 1
  352. }
  353. }
  354.  
  355. layer {
  356. name: "conv3"
  357. type: "Convolution"
  358. bottom: "res3_4"
  359. top: "conv3"
  360. param {
  361. lr_mult: 1
  362. decay_mult: 1
  363. }
  364. param {
  365. lr_mult: 2
  366. decay_mult: 0
  367. }
  368. convolution_param {
  369. num_output: 256
  370. kernel_size: 3
  371. stride: 1
  372. weight_filler {
  373. type: "xavier"
  374. }
  375. bias_filler {
  376. type: "constant"
  377. value: 0
  378. }
  379. }
  380. }
  381. layer {
  382. name: "relu3"
  383. type: "PReLU"
  384. bottom: "conv3"
  385. top: "conv3"
  386. }
  387. layer {
  388. name: "pool3"
  389. type: "Pooling"
  390. bottom: "conv3"
  391. top: "pool3"
  392. pooling_param {
  393. pool: MAX
  394. kernel_size: 2
  395. stride: 2
  396. }
  397. }
  398. layer {
  399. name: "conv4_1"
  400. type: "Convolution"
  401. bottom: "pool3"
  402. top: "conv4_1"
  403. param {
  404. lr_mult: 1
  405. decay_mult: 1
  406. }
  407. param {
  408. lr_mult: 0
  409. decay_mult: 0
  410. }
  411. convolution_param {
  412. num_output: 256
  413. kernel_size: 3
  414. stride: 1
  415. pad: 1
  416. weight_filler {
  417. type: "gaussian"
  418. std: 0.01
  419. }
  420. bias_filler {
  421. type: "constant"
  422. value: 0
  423. }
  424. }
  425. }
  426. layer {
  427. name: "relu4_1"
  428. type: "PReLU"
  429. bottom: "conv4_1"
  430. top: "conv4_1"
  431. }
  432. layer {
  433. name: "conv4_2"
  434. type: "Convolution"
  435. bottom: "conv4_1"
  436. top: "conv4_2"
  437. param {
  438. lr_mult: 1
  439. decay_mult: 1
  440. }
  441. param {
  442. lr_mult: 0
  443. decay_mult: 0
  444. }
  445. convolution_param {
  446. num_output: 256
  447. kernel_size: 3
  448. stride: 1
  449. pad: 1
  450. weight_filler {
  451. type: "gaussian"
  452. std: 0.01
  453. }
  454. bias_filler {
  455. type: "constant"
  456. value: 0
  457. }
  458. }
  459. }
  460. layer {
  461. name: "relu4_2"
  462. type: "PReLU"
  463. bottom: "conv4_2"
  464. top: "conv4_2"
  465. }
  466. layer {
  467. name: "res4_2"
  468. type: "Eltwise"
  469. bottom: "pool3"
  470. bottom: "conv4_2"
  471. top: "res4_2"
  472. eltwise_param {
  473. operation: 1
  474. }
  475. }
  476. layer {
  477. name: "conv4_3"
  478. type: "Convolution"
  479. bottom: "res4_2"
  480. top: "conv4_3"
  481. param {
  482. lr_mult: 1
  483. decay_mult: 1
  484. }
  485. param {
  486. lr_mult: 0
  487. decay_mult: 0
  488. }
  489. convolution_param {
  490. num_output: 256
  491. kernel_size: 3
  492. stride: 1
  493. pad: 1
  494. weight_filler {
  495. type: "gaussian"
  496. std: 0.01
  497. }
  498. bias_filler {
  499. type: "constant"
  500. value: 0
  501. }
  502. }
  503. }
  504. layer {
  505. name: "relu4_3"
  506. type: "PReLU"
  507. bottom: "conv4_3"
  508. top: "conv4_3"
  509. }
  510. layer {
  511. name: "conv4_4"
  512. type: "Convolution"
  513. bottom: "conv4_3"
  514. top: "conv4_4"
  515. param {
  516. lr_mult: 1
  517. decay_mult: 1
  518. }
  519. param {
  520. lr_mult: 0
  521. decay_mult: 0
  522. }
  523. convolution_param {
  524. num_output: 256
  525. kernel_size: 3
  526. stride: 1
  527. pad: 1
  528. weight_filler {
  529. type: "gaussian"
  530. std: 0.01
  531. }
  532. bias_filler {
  533. type: "constant"
  534. value: 0
  535. }
  536. }
  537. }
  538. layer {
  539. name: "relu4_4"
  540. type: "PReLU"
  541. bottom: "conv4_4"
  542. top: "conv4_4"
  543. }
  544. layer {
  545. name: "res4_4"
  546. type: "Eltwise"
  547. bottom: "res4_2"
  548. bottom: "conv4_4"
  549. top: "res4_4"
  550. eltwise_param {
  551. operation: 1
  552. }
  553. }
  554. layer {
  555. name: "conv4_5"
  556. type: "Convolution"
  557. bottom: "res4_4"
  558. top: "conv4_5"
  559. param {
  560. lr_mult: 1
  561. decay_mult: 1
  562. }
  563. param {
  564. lr_mult: 0
  565. decay_mult: 0
  566. }
  567. convolution_param {
  568. num_output: 256
  569. kernel_size: 3
  570. stride: 1
  571. pad: 1
  572. weight_filler {
  573. type: "gaussian"
  574. std: 0.01
  575. }
  576. bias_filler {
  577. type: "constant"
  578. value: 0
  579. }
  580. }
  581. }
  582. layer {
  583. name: "relu4_5"
  584. type: "PReLU"
  585. bottom: "conv4_5"
  586. top: "conv4_5"
  587. }
  588. layer {
  589. name: "conv4_6"
  590. type: "Convolution"
  591. bottom: "conv4_5"
  592. top: "conv4_6"
  593. param {
  594. lr_mult: 1
  595. decay_mult: 1
  596. }
  597. param {
  598. lr_mult: 0
  599. decay_mult: 0
  600. }
  601. convolution_param {
  602. num_output: 256
  603. kernel_size: 3
  604. stride: 1
  605. pad: 1
  606. weight_filler {
  607. type: "gaussian"
  608. std: 0.01
  609. }
  610. bias_filler {
  611. type: "constant"
  612. value: 0
  613. }
  614. }
  615. }
  616. layer {
  617. name: "relu4_6"
  618. type: "PReLU"
  619. bottom: "conv4_6"
  620. top: "conv4_6"
  621. }
  622. layer {
  623. name: "res4_6"
  624. type: "Eltwise"
  625. bottom: "res4_4"
  626. bottom: "conv4_6"
  627. top: "res4_6"
  628. eltwise_param {
  629. operation: 1
  630. }
  631. }
  632. layer {
  633. name: "conv4_7"
  634. type: "Convolution"
  635. bottom: "res4_6"
  636. top: "conv4_7"
  637. param {
  638. lr_mult: 1
  639. decay_mult: 1
  640. }
  641. param {
  642. lr_mult: 0
  643. decay_mult: 0
  644. }
  645. convolution_param {
  646. num_output: 256
  647. kernel_size: 3
  648. stride: 1
  649. pad: 1
  650. weight_filler {
  651. type: "gaussian"
  652. std: 0.01
  653. }
  654. bias_filler {
  655. type: "constant"
  656. value: 0
  657. }
  658. }
  659. }
  660. layer {
  661. name: "relu4_7"
  662. type: "PReLU"
  663. bottom: "conv4_7"
  664. top: "conv4_7"
  665. }
  666. layer {
  667. name: "conv4_8"
  668. type: "Convolution"
  669. bottom: "conv4_7"
  670. top: "conv4_8"
  671. param {
  672. lr_mult: 1
  673. decay_mult: 1
  674. }
  675. param {
  676. lr_mult: 0
  677. decay_mult: 0
  678. }
  679. convolution_param {
  680. num_output: 256
  681. kernel_size: 3
  682. stride: 1
  683. pad: 1
  684. weight_filler {
  685. type: "gaussian"
  686. std: 0.01
  687. }
  688. bias_filler {
  689. type: "constant"
  690. value: 0
  691. }
  692. }
  693. }
  694. layer {
  695. name: "relu4_8"
  696. type: "PReLU"
  697. bottom: "conv4_8"
  698. top: "conv4_8"
  699. }
  700. layer {
  701. name: "res4_8"
  702. type: "Eltwise"
  703. bottom: "res4_6"
  704. bottom: "conv4_8"
  705. top: "res4_8"
  706. eltwise_param {
  707. operation: 1
  708. }
  709. }
  710. layer {
  711. name: "conv4_9"
  712. type: "Convolution"
  713. bottom: "res4_8"
  714. top: "conv4_9"
  715. param {
  716. lr_mult: 1
  717. decay_mult: 1
  718. }
  719. param {
  720. lr_mult: 0
  721. decay_mult: 0
  722. }
  723. convolution_param {
  724. num_output: 256
  725. kernel_size: 3
  726. stride: 1
  727. pad: 1
  728. weight_filler {
  729. type: "gaussian"
  730. std: 0.01
  731. }
  732. bias_filler {
  733. type: "constant"
  734. value: 0
  735. }
  736. }
  737. }
  738. layer {
  739. name: "relu4_9"
  740. type: "PReLU"
  741. bottom: "conv4_9"
  742. top: "conv4_9"
  743. }
  744. layer {
  745. name: "conv4_10"
  746. type: "Convolution"
  747. bottom: "conv4_9"
  748. top: "conv4_10"
  749. param {
  750. lr_mult: 1
  751. decay_mult: 1
  752. }
  753. param {
  754. lr_mult: 0
  755. decay_mult: 0
  756. }
  757. convolution_param {
  758. num_output: 256
  759. kernel_size: 3
  760. stride: 1
  761. pad: 1
  762. weight_filler {
  763. type: "gaussian"
  764. std: 0.01
  765. }
  766. bias_filler {
  767. type: "constant"
  768. value: 0
  769. }
  770. }
  771. }
  772. layer {
  773. name: "relu4_10"
  774. type: "PReLU"
  775. bottom: "conv4_10"
  776. top: "conv4_10"
  777. }
  778. layer {
  779. name: "res4_10"
  780. type: "Eltwise"
  781. bottom: "res4_8"
  782. bottom: "conv4_10"
  783. top: "res4_10"
  784. eltwise_param {
  785. operation: 1
  786. }
  787. }
  788. layer {
  789. name: "conv4"
  790. type: "Convolution"
  791. bottom: "res4_10"
  792. top: "conv4"
  793. param {
  794. lr_mult: 1
  795. decay_mult: 1
  796. }
  797. param {
  798. lr_mult: 2
  799. decay_mult: 0
  800. }
  801. convolution_param {
  802. num_output: 512
  803. kernel_size: 3
  804. stride: 1
  805. weight_filler {
  806. type: "xavier"
  807. }
  808. bias_filler {
  809. type: "constant"
  810. value: 0
  811. }
  812. }
  813. }
  814. layer {
  815. name: "relu4"
  816. type: "PReLU"
  817. bottom: "conv4"
  818. top: "conv4"
  819. }
  820. layer {
  821. name: "pool4"
  822. type: "Pooling"
  823. bottom: "conv4"
  824. top: "pool4"
  825. pooling_param {
  826. pool: MAX
  827. kernel_size: 2
  828. stride: 2
  829. }
  830. }
  831. layer {
  832. name: "conv5_1"
  833. type: "Convolution"
  834. bottom: "pool4"
  835. top: "conv5_1"
  836. param {
  837. lr_mult: 1
  838. decay_mult: 1
  839. }
  840. param {
  841. lr_mult: 0
  842. decay_mult: 0
  843. }
  844. convolution_param {
  845. num_output: 512
  846. kernel_size: 3
  847. stride: 1
  848. pad: 1
  849. weight_filler {
  850. type: "gaussian"
  851. std: 0.01
  852. }
  853. bias_filler {
  854. type: "constant"
  855. value: 0
  856. }
  857. }
  858. }
  859. layer {
  860. name: "relu5_1"
  861. type: "PReLU"
  862. bottom: "conv5_1"
  863. top: "conv5_1"
  864. }
  865. layer {
  866. name: "conv5_2"
  867. type: "Convolution"
  868. bottom: "conv5_1"
  869. top: "conv5_2"
  870. param {
  871. lr_mult: 1
  872. decay_mult: 1
  873. }
  874. param {
  875. lr_mult: 0
  876. decay_mult: 0
  877. }
  878. convolution_param {
  879. num_output: 512
  880. kernel_size: 3
  881. stride: 1
  882. pad: 1
  883. weight_filler {
  884. type: "gaussian"
  885. std: 0.01
  886. }
  887. bias_filler {
  888. type: "constant"
  889. value: 0
  890. }
  891. }
  892. }
  893. layer {
  894. name: "relu5_2"
  895. type: "PReLU"
  896. bottom: "conv5_2"
  897. top: "conv5_2"
  898. }
  899. layer {
  900. name: "res5_2"
  901. type: "Eltwise"
  902. bottom: "pool4"
  903. bottom: "conv5_2"
  904. top: "res5_2"
  905. eltwise_param {
  906. operation: 1
  907. }
  908. }
  909. layer {
  910. name: "conv5_3"
  911. type: "Convolution"
  912. bottom: "res5_2"
  913. top: "conv5_3"
  914. param {
  915. lr_mult: 1
  916. decay_mult: 1
  917. }
  918. param {
  919. lr_mult: 0
  920. decay_mult: 0
  921. }
  922. convolution_param {
  923. num_output: 512
  924. kernel_size: 3
  925. stride: 1
  926. pad: 1
  927. weight_filler {
  928. type: "gaussian"
  929. std: 0.01
  930. }
  931. bias_filler {
  932. type: "constant"
  933. value: 0
  934. }
  935. }
  936. }
  937. layer {
  938. name: "relu5_3"
  939. type: "PReLU"
  940. bottom: "conv5_3"
  941. top: "conv5_3"
  942. }
  943. layer {
  944. name: "conv5_4"
  945. type: "Convolution"
  946. bottom: "conv5_3"
  947. top: "conv5_4"
  948. param {
  949. lr_mult: 1
  950. decay_mult: 1
  951. }
  952. param {
  953. lr_mult: 0
  954. decay_mult: 0
  955. }
  956. convolution_param {
  957. num_output: 512
  958. kernel_size: 3
  959. stride: 1
  960. pad: 1
  961. weight_filler {
  962. type: "gaussian"
  963. std: 0.01
  964. }
  965. bias_filler {
  966. type: "constant"
  967. value: 0
  968. }
  969. }
  970. }
  971. layer {
  972. name: "relu5_4"
  973. type: "PReLU"
  974. bottom: "conv5_4"
  975. top: "conv5_4"
  976. }
  977. layer {
  978. name: "res5_4"
  979. type: "Eltwise"
  980. bottom: "res5_2"
  981. bottom: "conv5_4"
  982. top: "res5_4"
  983. eltwise_param {
  984. operation: 1
  985. }
  986. }
  987. layer {
  988. name: "conv5_5"
  989. type: "Convolution"
  990. bottom: "res5_4"
  991. top: "conv5_5"
  992. param {
  993. lr_mult: 1
  994. decay_mult: 1
  995. }
  996. param {
  997. lr_mult: 0
  998. decay_mult: 0
  999. }
  1000. convolution_param {
  1001. num_output: 512
  1002. kernel_size: 3
  1003. stride: 1
  1004. pad: 1
  1005. weight_filler {
  1006. type: "gaussian"
  1007. std: 0.01
  1008. }
  1009. bias_filler {
  1010. type: "constant"
  1011. value: 0
  1012. }
  1013. }
  1014. }
  1015. layer {
  1016. name: "relu5_5"
  1017. type: "PReLU"
  1018. bottom: "conv5_5"
  1019. top: "conv5_5"
  1020. }
  1021. layer {
  1022. name: "conv5_6"
  1023. type: "Convolution"
  1024. bottom: "conv5_5"
  1025. top: "conv5_6"
  1026. param {
  1027. lr_mult: 1
  1028. decay_mult: 1
  1029. }
  1030. param {
  1031. lr_mult: 0
  1032. decay_mult: 0
  1033. }
  1034. convolution_param {
  1035. num_output: 512
  1036. kernel_size: 3
  1037. stride: 1
  1038. pad: 1
  1039. weight_filler {
  1040. type: "gaussian"
  1041. std: 0.01
  1042. }
  1043. bias_filler {
  1044. type: "constant"
  1045. value: 0
  1046. }
  1047. }
  1048. }
  1049. layer {
  1050. name: "relu5_6"
  1051. type: "PReLU"
  1052. bottom: "conv5_6"
  1053. top: "conv5_6"
  1054. }
  1055. layer {
  1056. name: "res5_6"
  1057. type: "Eltwise"
  1058. bottom: "res5_4"
  1059. bottom: "conv5_6"
  1060. top: "res5_6"
  1061. eltwise_param {
  1062. operation: 1
  1063. }
  1064. }
  1065. layer {
  1066. name: "fc5"
  1067. type: "InnerProduct"
  1068. bottom: "res5_6"
  1069. top: "fc5"
  1070. param {
  1071. lr_mult: 1
  1072. decay_mult: 1
  1073. }
  1074. param {
  1075. lr_mult: 2
  1076. decay_mult: 0
  1077. }
  1078. inner_product_param {
  1079. num_output: 512
  1080. weight_filler {
  1081. type: "xavier"
  1082. }
  1083. bias_filler {
  1084. type: "constant"
  1085. value: 0
  1086. }
  1087. }
  1088. }
  1089. ############## softmax loss ###############
  1090. layer {
  1091. name: "fc6"
  1092. type: "InnerProduct"
  1093. bottom: "fc5"
  1094. top: "fc6"
  1095. param {
  1096. lr_mult: 1
  1097. decay_mult: 1
  1098. }
  1099. param {
  1100. lr_mult: 2
  1101. decay_mult: 0
  1102. }
  1103. inner_product_param {
  1104. num_output: 10572
  1105. weight_filler {
  1106. type: "xavier"
  1107. }
  1108. bias_filler {
  1109. type: "constant"
  1110. value: 0
  1111. }
  1112. }
  1113. }
  1114. layer {
  1115. name: "softmax_loss"
  1116. type: "SoftmaxWithLoss"
  1117. bottom: "fc6"
  1118. bottom: "label"
  1119. top: "softmax_loss"
  1120. }
  1121. ############## center loss ###############
  1122. layer {
  1123. name: "center_loss"
  1124. type: "CenterLoss"
  1125. bottom: "fc5"
  1126. bottom: "label"
  1127. top: "center_loss"
  1128. param {
  1129. lr_mult: 1
  1130. decay_mult: 2
  1131. }
  1132. center_loss_param {
  1133. num_output: 10572
  1134. center_filler {
  1135. type: "xavier"
  1136. }
  1137. }
  1138. loss_weight: 0.008
  1139. }
Add Comment
Please, Sign In to add comment
Public Pastes
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!