API tools faq
paste
Advertisement
Guest User

Untitled

a guest
Nov 20th, 2019
126
0
Never
Add comment
Not a member of Pastebin yet? Sign Up, it unlocks many cool features!
text 15.16 KB | None | 0 0
raw download clone embed print report
  1. Configurations:
  2. BACKBONE resnet101
  3. BACKBONE_STRIDES [4, 8, 16, 32, 64]
  4. BATCH_SIZE 4
  5. BBOX_STD_DEV [0.1 0.1 0.2 0.2]
  6. COMPUTE_BACKBONE_SHAPE None
  7. DETECTION_MAX_INSTANCES 100
  8. DETECTION_MIN_CONFIDENCE 0.9
  9. DETECTION_NMS_THRESHOLD 0.3
  10. FPN_CLASSIF_FC_LAYERS_SIZE 1024
  11. GPU_COUNT 2
  12. GRADIENT_CLIP_NORM 5.0
  13. IMAGES_PER_GPU 2
  14. IMAGE_CHANNEL_COUNT 3
  15. IMAGE_MAX_DIM 1024
  16. IMAGE_META_SIZE 19
  17. IMAGE_MIN_DIM 800
  18. IMAGE_MIN_SCALE 0
  19. IMAGE_RESIZE_MODE square
  20. IMAGE_SHAPE [1024 1024 3]
  21. LEARNING_MOMENTUM 0.9
  22. LEARNING_RATE 0.001
  23. LOSS_WEIGHTS {'rpn_class_loss': 1.0, 'rpn_bbox_loss': 1.0, 'mrcnn_class_loss': 1.0, 'mrcnn_bbox_loss': 1.0, 'mrcnn_mask_loss': 1.0}
  24. MASK_POOL_SIZE 14
  25. MASK_SHAPE [28, 28]
  26. MAX_GT_INSTANCES 100
  27. MEAN_PIXEL [123.7 116.8 103.9]
  28. MINI_MASK_SHAPE (56, 56)
  29. NAME ABBP
  30. NUM_CLASSES 7
  31. POOL_SIZE 7
  32. POST_NMS_ROIS_INFERENCE 1000
  33. POST_NMS_ROIS_TRAINING 2000
  34. PRE_NMS_LIMIT 6000
  35. ROI_POSITIVE_RATIO 0.33
  36. RPN_ANCHOR_RATIOS [0.5, 1, 2]
  37. RPN_ANCHOR_SCALES (32, 64, 128, 256, 512)
  38. RPN_ANCHOR_STRIDE 1
  39. RPN_BBOX_STD_DEV [0.1 0.1 0.2 0.2]
  40. RPN_NMS_THRESHOLD 0.7
  41. RPN_TRAIN_ANCHORS_PER_IMAGE 256
  42. STEPS_PER_EPOCH 100
  43. TOP_DOWN_PYRAMID_SIZE 256
  44. TRAIN_BN False
  45. TRAIN_ROIS_PER_IMAGE 200
  46. USE_MINI_MASK True
  47. USE_RPN_ROIS True
  48. VALIDATION_STEPS 50
  49. WEIGHT_DECAY 0.0001 Configurations:
  50. BACKBONE resnet101
  51. BACKBONE_STRIDES [4, 8, 16, 32, 64]
  52. BATCH_SIZE 4
  53. BBOX_STD_DEV [0.1 0.1 0.2 0.2]
  54. COMPUTE_BACKBONE_SHAPE None
  55. DETECTION_MAX_INSTANCES 100
  56. DETECTION_MIN_CONFIDENCE 0.9
  57. DETECTION_NMS_THRESHOLD 0.3
  58. FPN_CLASSIF_FC_LAYERS_SIZE 1024
  59. GPU_COUNT 2
  60. GRADIENT_CLIP_NORM 5.0
  61. IMAGES_PER_GPU 2
  62. IMAGE_CHANNEL_COUNT 3
  63. IMAGE_MAX_DIM 1024
  64. IMAGE_META_SIZE 19
  65. IMAGE_MIN_DIM 800
  66. IMAGE_MIN_SCALE 0
  67. IMAGE_RESIZE_MODE square
  68. IMAGE_SHAPE [1024 1024 3]
  69. LEARNING_MOMENTUM 0.9
  70. LEARNING_RATE 0.001
  71. LOSS_WEIGHTS {'rpn_class_loss': 1.0, 'rpn_bbox_loss': 1.0, 'mrcnn_class_loss': 1.0, 'mrcnn_bbox_loss': 1.0, 'mrcnn_mask_loss': 1.0}
  72. MASK_POOL_SIZE 14
  73. MASK_SHAPE [28, 28]
  74. MAX_GT_INSTANCES 100
  75. MEAN_PIXEL [123.7 116.8 103.9]
  76. MINI_MASK_SHAPE (56, 56)
  77. NAME ABBP
  78. NUM_CLASSES 7
  79. POOL_SIZE 7
  80. POST_NMS_ROIS_INFERENCE 1000
  81. POST_NMS_ROIS_TRAINING 2000
  82. PRE_NMS_LIMIT 6000
  83. ROI_POSITIVE_RATIO 0.33
  84. RPN_ANCHOR_RATIOS [0.5, 1, 2]
  85. RPN_ANCHOR_SCALES (32, 64, 128, 256, 512)
  86. RPN_ANCHOR_STRIDE 1
  87. RPN_BBOX_STD_DEV [0.1 0.1 0.2 0.2]
  88. RPN_NMS_THRESHOLD 0.7
  89. RPN_TRAIN_ANCHORS_PER_IMAGE 256
  90. STEPS_PER_EPOCH 100
  91. TOP_DOWN_PYRAMID_SIZE 256
  92. TRAIN_BN False
  93. TRAIN_ROIS_PER_IMAGE 200
  94. USE_MINI_MASK True
  95. USE_RPN_ROIS True
  96. VALIDATION_STEPS 50
  97. WEIGHT_DECAY 0.0001
  98.  
  99. LR 0.002
  100.  
  101. Epoch 1/30
  102. 100/100 [==============================] - 219s 2s/step - loss: 0.7387 - rpn_class_loss: 0.0027 - rpn_bbox_loss: 0.1275 - mrcnn_class_loss: 0.1527 - mrcnn_bbox_loss: 0.2608 - mrcnn_mask_loss: 0.1949 - val_loss: 0.4981 - val_rpn_class_loss: 0.0041 - val_rpn_bbox_loss: 0.1694 - val_mrcnn_class_loss: 0.0716 - val_mrcnn_bbox_loss: 0.1824 - val_mrcnn_mask_loss: 0.0706
  103. WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/callbacks.py:1265: The name tf.Summary is deprecated. Please use tf.compat.v1.Summary instead.
  104.  
  105. Epoch 2/30
  106. 100/100 [==============================] - 134s 1s/step - loss: 0.4390 - rpn_class_loss: 0.0031 - rpn_bbox_loss: 0.1459 - mrcnn_class_loss: 0.0831 - mrcnn_bbox_loss: 0.1461 - mrcnn_mask_loss: 0.0608 - val_loss: 0.5152 - val_rpn_class_loss: 0.0035 - val_rpn_bbox_loss: 0.2463 - val_mrcnn_class_loss: 0.0565 - val_mrcnn_bbox_loss: 0.1534 - val_mrcnn_mask_loss: 0.0555
  107. Epoch 3/30
  108. 100/100 [==============================] - 134s 1s/step - loss: 0.4096 - rpn_class_loss: 0.0038 - rpn_bbox_loss: 0.1541 - mrcnn_class_loss: 0.0695 - mrcnn_bbox_loss: 0.1236 - mrcnn_mask_loss: 0.0586 - val_loss: 0.4742 - val_rpn_class_loss: 0.0037 - val_rpn_bbox_loss: 0.1648 - val_mrcnn_class_loss: 0.0646 - val_mrcnn_bbox_loss: 0.1253 - val_mrcnn_mask_loss: 0.1158
  109. Epoch 4/30
  110. 100/100 [==============================] - 134s 1s/step - loss: 0.3657 - rpn_class_loss: 0.0038 - rpn_bbox_loss: 0.1459 - mrcnn_class_loss: 0.0574 - mrcnn_bbox_loss: 0.0843 - mrcnn_mask_loss: 0.0744 - val_loss: 0.5684 - val_rpn_class_loss: 0.0045 - val_rpn_bbox_loss: 0.2482 - val_mrcnn_class_loss: 0.0852 - val_mrcnn_bbox_loss: 0.1595 - val_mrcnn_mask_loss: 0.0709
  111. Epoch 5/30
  112. 100/100 [==============================] - 134s 1s/step - loss: 0.3664 - rpn_class_loss: 0.0027 - rpn_bbox_loss: 0.1460 - mrcnn_class_loss: 0.0593 - mrcnn_bbox_loss: 0.0816 - mrcnn_mask_loss: 0.0769 - val_loss: 0.4589 - val_rpn_class_loss: 0.0037 - val_rpn_bbox_loss: 0.2420 - val_mrcnn_class_loss: 0.0521 - val_mrcnn_bbox_loss: 0.0969 - val_mrcnn_mask_loss: 0.0642
  113. Epoch 6/30
  114. 100/100 [==============================] - 134s 1s/step - loss: 0.4407 - rpn_class_loss: 0.0090 - rpn_bbox_loss: 0.2534 - mrcnn_class_loss: 0.0408 - mrcnn_bbox_loss: 0.0851 - mrcnn_mask_loss: 0.0523 - val_loss: 0.4750 - val_rpn_class_loss: 0.0039 - val_rpn_bbox_loss: 0.1853 - val_mrcnn_class_loss: 0.0582 - val_mrcnn_bbox_loss: 0.1547 - val_mrcnn_mask_loss: 0.0728
  115. Epoch 7/30
  116. 100/100 [==============================] - 133s 1s/step - loss: 0.3703 - rpn_class_loss: 0.0037 - rpn_bbox_loss: 0.1591 - mrcnn_class_loss: 0.0437 - mrcnn_bbox_loss: 0.0929 - mrcnn_mask_loss: 0.0708 - val_loss: 0.3630 - val_rpn_class_loss: 0.0037 - val_rpn_bbox_loss: 0.1542 - val_mrcnn_class_loss: 0.0406 - val_mrcnn_bbox_loss: 0.0982 - val_mrcnn_mask_loss: 0.0663
  117. Epoch 8/30
  118. 100/100 [==============================] - 134s 1s/step - loss: 0.3879 - rpn_class_loss: 0.0035 - rpn_bbox_loss: 0.1692 - mrcnn_class_loss: 0.0500 - mrcnn_bbox_loss: 0.0770 - mrcnn_mask_loss: 0.0882 - val_loss: 0.4490 - val_rpn_class_loss: 0.0029 - val_rpn_bbox_loss: 0.2095 - val_mrcnn_class_loss: 0.0449 - val_mrcnn_bbox_loss: 0.1326 - val_mrcnn_mask_loss: 0.0591
  119. Epoch 9/30
  120. 100/100 [==============================] - 133s 1s/step - loss: 0.3590 - rpn_class_loss: 0.0030 - rpn_bbox_loss: 0.1787 - mrcnn_class_loss: 0.0409 - mrcnn_bbox_loss: 0.0650 - mrcnn_mask_loss: 0.0713 - val_loss: 0.3113 - val_rpn_class_loss: 0.0013 - val_rpn_bbox_loss: 0.1644 - val_mrcnn_class_loss: 0.0267 - val_mrcnn_bbox_loss: 0.0614 - val_mrcnn_mask_loss: 0.0575
  121. Epoch 10/30
  122. 100/100 [==============================] - 134s 1s/step - loss: 0.3312 - rpn_class_loss: 0.0035 - rpn_bbox_loss: 0.1527 - mrcnn_class_loss: 0.0362 - mrcnn_bbox_loss: 0.0669 - mrcnn_mask_loss: 0.0719 - val_loss: 0.4589 - val_rpn_class_loss: 0.0047 - val_rpn_bbox_loss: 0.2388 - val_mrcnn_class_loss: 0.0438 - val_mrcnn_bbox_loss: 0.0865 - val_mrcnn_mask_loss: 0.0850
  123. Epoch 11/30
  124. 100/100 [==============================] - 134s 1s/step - loss: 0.2840 - rpn_class_loss: 0.0031 - rpn_bbox_loss: 0.1222 - mrcnn_class_loss: 0.0360 - mrcnn_bbox_loss: 0.0498 - mrcnn_mask_loss: 0.0729 - val_loss: 0.3430 - val_rpn_class_loss: 0.0038 - val_rpn_bbox_loss: 0.1646 - val_mrcnn_class_loss: 0.0484 - val_mrcnn_bbox_loss: 0.0727 - val_mrcnn_mask_loss: 0.0535
  125. Epoch 12/30
  126. 100/100 [==============================] - 135s 1s/step - loss: 0.3001 - rpn_class_loss: 0.0034 - rpn_bbox_loss: 0.1567 - mrcnn_class_loss: 0.0286 - mrcnn_bbox_loss: 0.0615 - mrcnn_mask_loss: 0.0499 - val_loss: 0.3364 - val_rpn_class_loss: 0.0040 - val_rpn_bbox_loss: 0.1637 - val_mrcnn_class_loss: 0.0360 - val_mrcnn_bbox_loss: 0.0778 - val_mrcnn_mask_loss: 0.0548
  127. Epoch 13/30
  128. 100/100 [==============================] - 134s 1s/step - loss: 0.3217 - rpn_class_loss: 0.0026 - rpn_bbox_loss: 0.1614 - mrcnn_class_loss: 0.0287 - mrcnn_bbox_loss: 0.0592 - mrcnn_mask_loss: 0.0697 - val_loss: 0.3600 - val_rpn_class_loss: 0.0026 - val_rpn_bbox_loss: 0.2057 - val_mrcnn_class_loss: 0.0219 - val_mrcnn_bbox_loss: 0.0708 - val_mrcnn_mask_loss: 0.0590
  129. Epoch 14/30
  130. 100/100 [==============================] - 134s 1s/step - loss: 0.2782 - rpn_class_loss: 0.0026 - rpn_bbox_loss: 0.1382 - mrcnn_class_loss: 0.0224 - mrcnn_bbox_loss: 0.0496 - mrcnn_mask_loss: 0.0654 - val_loss: 0.3160 - val_rpn_class_loss: 0.0028 - val_rpn_bbox_loss: 0.1601 - val_mrcnn_class_loss: 0.0270 - val_mrcnn_bbox_loss: 0.0638 - val_mrcnn_mask_loss: 0.0622
  131. Epoch 15/30
  132. 100/100 [==============================] - 134s 1s/step - loss: 0.3010 - rpn_class_loss: 0.0025 - rpn_bbox_loss: 0.1516 - mrcnn_class_loss: 0.0240 - mrcnn_bbox_loss: 0.0545 - mrcnn_mask_loss: 0.0683 - val_loss: 0.3311 - val_rpn_class_loss: 0.0025 - val_rpn_bbox_loss: 0.1575 - val_mrcnn_class_loss: 0.0352 - val_mrcnn_bbox_loss: 0.0769 - val_mrcnn_mask_loss: 0.0591
  133. Epoch 16/30
  134. 100/100 [==============================] - 133s 1s/step - loss: 0.2883 - rpn_class_loss: 0.0025 - rpn_bbox_loss: 0.1508 - mrcnn_class_loss: 0.0301 - mrcnn_bbox_loss: 0.0487 - mrcnn_mask_loss: 0.0562 - val_loss: 0.3576 - val_rpn_class_loss: 0.0031 - val_rpn_bbox_loss: 0.1834 - val_mrcnn_class_loss: 0.0283 - val_mrcnn_bbox_loss: 0.0728 - val_mrcnn_mask_loss: 0.0700
  135. Epoch 17/30
  136. 100/100 [==============================] - 134s 1s/step - loss: 0.2457 - rpn_class_loss: 0.0030 - rpn_bbox_loss: 0.1216 - mrcnn_class_loss: 0.0248 - mrcnn_bbox_loss: 0.0476 - mrcnn_mask_loss: 0.0488 - val_loss: 0.2930 - val_rpn_class_loss: 0.0031 - val_rpn_bbox_loss: 0.1623 - val_mrcnn_class_loss: 0.0287 - val_mrcnn_bbox_loss: 0.0402 - val_mrcnn_mask_loss: 0.0587
  137. Epoch 18/30
  138. 100/100 [==============================] - 134s 1s/step - loss: 0.2668 - rpn_class_loss: 0.0034 - rpn_bbox_loss: 0.1211 - mrcnn_class_loss: 0.0444 - mrcnn_bbox_loss: 0.0498 - mrcnn_mask_loss: 0.0482 - val_loss: 0.3211 - val_rpn_class_loss: 0.0042 - val_rpn_bbox_loss: 0.1752 - val_mrcnn_class_loss: 0.0219 - val_mrcnn_bbox_loss: 0.0592 - val_mrcnn_mask_loss: 0.0607
  139. Epoch 19/30
  140. 100/100 [==============================] - 134s 1s/step - loss: 0.2336 - rpn_class_loss: 0.0019 - rpn_bbox_loss: 0.1135 - mrcnn_class_loss: 0.0251 - mrcnn_bbox_loss: 0.0421 - mrcnn_mask_loss: 0.0510 - val_loss: 0.3061 - val_rpn_class_loss: 0.0012 - val_rpn_bbox_loss: 0.1951 - val_mrcnn_class_loss: 0.0149 - val_mrcnn_bbox_loss: 0.0418 - val_mrcnn_mask_loss: 0.0532
  141. Epoch 20/30
  142. 100/100 [==============================] - 134s 1s/step - loss: 0.2444 - rpn_class_loss: 0.0018 - rpn_bbox_loss: 0.1300 - mrcnn_class_loss: 0.0214 - mrcnn_bbox_loss: 0.0415 - mrcnn_mask_loss: 0.0497 - val_loss: 0.3741 - val_rpn_class_loss: 0.0012 - val_rpn_bbox_loss: 0.2214 - val_mrcnn_class_loss: 0.0485 - val_mrcnn_bbox_loss: 0.0518 - val_mrcnn_mask_loss: 0.0512
  143. Epoch 21/30
  144. 100/100 [==============================] - 134s 1s/step - loss: 0.2443 - rpn_class_loss: 0.0019 - rpn_bbox_loss: 0.1269 - mrcnn_class_loss: 0.0270 - mrcnn_bbox_loss: 0.0387 - mrcnn_mask_loss: 0.0498 - val_loss: 0.3579 - val_rpn_class_loss: 0.0048 - val_rpn_bbox_loss: 0.1901 - val_mrcnn_class_loss: 0.0247 - val_mrcnn_bbox_loss: 0.0583 - val_mrcnn_mask_loss: 0.0799
  145. Epoch 22/30
  146. 100/100 [==============================] - 134s 1s/step - loss: 0.1686 - rpn_class_loss: 0.0021 - rpn_bbox_loss: 0.0749 - mrcnn_class_loss: 0.0156 - mrcnn_bbox_loss: 0.0307 - mrcnn_mask_loss: 0.0453 - val_loss: 0.2296 - val_rpn_class_loss: 0.0021 - val_rpn_bbox_loss: 0.1253 - val_mrcnn_class_loss: 0.0125 - val_mrcnn_bbox_loss: 0.0465 - val_mrcnn_mask_loss: 0.0432
  147. Epoch 23/30
  148. 100/100 [==============================] - 134s 1s/step - loss: 0.1975 - rpn_class_loss: 0.0017 - rpn_bbox_loss: 0.0747 - mrcnn_class_loss: 0.0134 - mrcnn_bbox_loss: 0.0308 - mrcnn_mask_loss: 0.0769 - val_loss: 0.1994 - val_rpn_class_loss: 0.0018 - val_rpn_bbox_loss: 0.0857 - val_mrcnn_class_loss: 0.0172 - val_mrcnn_bbox_loss: 0.0418 - val_mrcnn_mask_loss: 0.0528
  149. Epoch 24/30
  150. 100/100 [==============================] - 134s 1s/step - loss: 0.2696 - rpn_class_loss: 0.0034 - rpn_bbox_loss: 0.1668 - mrcnn_class_loss: 0.0154 - mrcnn_bbox_loss: 0.0347 - mrcnn_mask_loss: 0.0493 - val_loss: 0.2640 - val_rpn_class_loss: 0.0027 - val_rpn_bbox_loss: 0.1037 - val_mrcnn_class_loss: 0.0323 - val_mrcnn_bbox_loss: 0.0471 - val_mrcnn_mask_loss: 0.0782
  151. Epoch 25/30
  152. 100/100 [==============================] - 133s 1s/step - loss: 0.2561 - rpn_class_loss: 0.0028 - rpn_bbox_loss: 0.1058 - mrcnn_class_loss: 0.0233 - mrcnn_bbox_loss: 0.0417 - mrcnn_mask_loss: 0.0826 - val_loss: 0.2262 - val_rpn_class_loss: 0.0015 - val_rpn_bbox_loss: 0.1035 - val_mrcnn_class_loss: 0.0135 - val_mrcnn_bbox_loss: 0.0448 - val_mrcnn_mask_loss: 0.0629
  153. Epoch 26/30
  154. 100/100 [==============================] - 132s 1s/step - loss: 0.2389 - rpn_class_loss: 0.0023 - rpn_bbox_loss: 0.1203 - mrcnn_class_loss: 0.0138 - mrcnn_bbox_loss: 0.0353 - mrcnn_mask_loss: 0.0672 - val_loss: 0.3307 - val_rpn_class_loss: 0.0034 - val_rpn_bbox_loss: 0.1749 - val_mrcnn_class_loss: 0.0299 - val_mrcnn_bbox_loss: 0.0452 - val_mrcnn_mask_loss: 0.0773
  155. Epoch 27/30
  156. 100/100 [==============================] - 133s 1s/step - loss: 0.2554 - rpn_class_loss: 0.0021 - rpn_bbox_loss: 0.1284 - mrcnn_class_loss: 0.0153 - mrcnn_bbox_loss: 0.0379 - mrcnn_mask_loss: 0.0718 - val_loss: 0.2728 - val_rpn_class_loss: 0.0031 - val_rpn_bbox_loss: 0.1687 - val_mrcnn_class_loss: 0.0104 - val_mrcnn_bbox_loss: 0.0355 - val_mrcnn_mask_loss: 0.0552
  157. Epoch 28/30
  158. 100/100 [==============================] - 133s 1s/step - loss: 0.2251 - rpn_class_loss: 0.0022 - rpn_bbox_loss: 0.1099 - mrcnn_class_loss: 0.0146 - mrcnn_bbox_loss: 0.0293 - mrcnn_mask_loss: 0.0690 - val_loss: 0.2814 - val_rpn_class_loss: 0.0020 - val_rpn_bbox_loss: 0.1774 - val_mrcnn_class_loss: 0.0130 - val_mrcnn_bbox_loss: 0.0299 - val_mrcnn_mask_loss: 0.0591
  159. Epoch 29/30
  160. 100/100 [==============================] - 133s 1s/step - loss: 0.1938 - rpn_class_loss: 0.0021 - rpn_bbox_loss: 0.0863 - mrcnn_class_loss: 0.0152 - mrcnn_bbox_loss: 0.0324 - mrcnn_mask_loss: 0.0578 - val_loss: 0.2685 - val_rpn_class_loss: 0.0016 - val_rpn_bbox_loss: 0.1530 - val_mrcnn_class_loss: 0.0168 - val_mrcnn_bbox_loss: 0.0455 - val_mrcnn_mask_loss: 0.0515
  161. Epoch 30/30
  162. 100/100 [==============================] - 133s 1s/step - loss: 0.2010 - rpn_class_loss: 0.0019 - rpn_bbox_loss: 0.1025 - mrcnn_class_loss: 0.0126 - mrcnn_bbox_loss: 0.0302 - mrcnn_mask_loss: 0.0538 - val_loss: 0.2368 - val_rpn_class_loss: 0.0019 - val_rpn_bbox_loss: 0.1189 - val_mrcnn_class_loss: 0.0106 - val_mrcnn_bbox_loss: 0.0507 - val_mrcnn_mask_loss: 0.0546
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!