search

SysCV / qdtrack

Quasi-Dense Similarity Learning for Multiple Object Tracking, CVPR 2021 (Oral)
Apache License 2.0
380 stars 61 forks source link

Provided checkpoint does NOT have weights for the Embedding head #138

Open tanjary21 opened 1 year ago

tanjary21 commented 1 year ago

Hi there.

Thank you for sharing such a powerful Detector-Tracker.

I noticed that the checkpoint you provide does not have pretrained weights for the embeddings head:

I printed out the keys of the state dict of the checkpoint you provide(first list) and the newly initialized model(second list). Notice how the embedding head is absent on the second list:

First List (keys of state_dict from provided ckpt):

  1. 'backbone.conv1.weight'
  2. 'backbone.bn1.weight'
  3. 'backbone.bn1.bias'
  4. 'backbone.bn1.running_mean'
  5. 'backbone.bn1.running_var'
  6. 'backbone.bn1.num_batches_tracked'
  7. 'backbone.layer1.0.conv1.weight'
  8. 'backbone.layer1.0.bn1.weight'
  9. 'backbone.layer1.0.bn1.bias'
  10. 'backbone.layer1.0.bn1.running_mean'
  11. 'backbone.layer1.0.bn1.running_var'
  12. 'backbone.layer1.0.bn1.num_batches_tracked'
  13. 'backbone.layer1.0.conv2.weight'
  14. 'backbone.layer1.0.bn2.weight'
  15. 'backbone.layer1.0.bn2.bias'
  16. 'backbone.layer1.0.bn2.running_mean'
  17. 'backbone.layer1.0.bn2.running_var'
  18. 'backbone.layer1.0.bn2.num_batches_tracked'
  19. 'backbone.layer1.0.conv3.weight'
  20. 'backbone.layer1.0.bn3.weight'
  21. 'backbone.layer1.0.bn3.bias'
  22. 'backbone.layer1.0.bn3.running_mean'
  23. 'backbone.layer1.0.bn3.running_var'
  24. 'backbone.layer1.0.bn3.num_batches_tracked'
  25. 'backbone.layer1.0.downsample.0.weight'
  26. 'backbone.layer1.0.downsample.1.weight'
  27. 'backbone.layer1.0.downsample.1.bias'
  28. 'backbone.layer1.0.downsample.1.running_mean'
  29. 'backbone.layer1.0.downsample.1.running_var'
  30. 'backbone.layer1.0.downsample.1.num_batches_tracked'
  31. 'backbone.layer1.1.conv1.weight'
  32. 'backbone.layer1.1.bn1.weight'
  33. 'backbone.layer1.1.bn1.bias'
  34. 'backbone.layer1.1.bn1.running_mean'
  35. 'backbone.layer1.1.bn1.running_var'
  36. 'backbone.layer1.1.bn1.num_batches_tracked'
  37. 'backbone.layer1.1.conv2.weight'
  38. 'backbone.layer1.1.bn2.weight'
  39. 'backbone.layer1.1.bn2.bias'
  40. 'backbone.layer1.1.bn2.running_mean'
  41. 'backbone.layer1.1.bn2.running_var'
  42. 'backbone.layer1.1.bn2.num_batches_tracked'
  43. 'backbone.layer1.1.conv3.weight'
  44. 'backbone.layer1.1.bn3.weight'
  45. 'backbone.layer1.1.bn3.bias'
  46. 'backbone.layer1.1.bn3.running_mean'
  47. 'backbone.layer1.1.bn3.running_var'
  48. 'backbone.layer1.1.bn3.num_batches_tracked'
  49. 'backbone.layer1.2.conv1.weight'
  50. 'backbone.layer1.2.bn1.weight'
  51. 'backbone.layer1.2.bn1.bias'
  52. 'backbone.layer1.2.bn1.running_mean'
  53. 'backbone.layer1.2.bn1.running_var'
  54. 'backbone.layer1.2.bn1.num_batches_tracked'
  55. 'backbone.layer1.2.conv2.weight'
  56. 'backbone.layer1.2.bn2.weight'
  57. 'backbone.layer1.2.bn2.bias'
  58. 'backbone.layer1.2.bn2.running_mean'
  59. 'backbone.layer1.2.bn2.running_var'
  60. 'backbone.layer1.2.bn2.num_batches_tracked'
  61. 'backbone.layer1.2.conv3.weight'
  62. 'backbone.layer1.2.bn3.weight'
  63. 'backbone.layer1.2.bn3.bias'
  64. 'backbone.layer1.2.bn3.running_mean'
  65. 'backbone.layer1.2.bn3.running_var'
  66. 'backbone.layer1.2.bn3.num_batches_tracked'
  67. 'backbone.layer2.0.conv1.weight'
  68. 'backbone.layer2.0.bn1.weight'
  69. 'backbone.layer2.0.bn1.bias'
  70. 'backbone.layer2.0.bn1.running_mean'
  71. 'backbone.layer2.0.bn1.running_var'
  72. 'backbone.layer2.0.bn1.num_batches_tracked'
  73. 'backbone.layer2.0.conv2.weight'
  74. 'backbone.layer2.0.bn2.weight'
  75. 'backbone.layer2.0.bn2.bias'
  76. 'backbone.layer2.0.bn2.running_mean'
  77. 'backbone.layer2.0.bn2.running_var'
  78. 'backbone.layer2.0.bn2.num_batches_tracked'
  79. 'backbone.layer2.0.conv3.weight'
  80. 'backbone.layer2.0.bn3.weight'
  81. 'backbone.layer2.0.bn3.bias'
  82. 'backbone.layer2.0.bn3.running_mean'
  83. 'backbone.layer2.0.bn3.running_var'
  84. 'backbone.layer2.0.bn3.num_batches_tracked'
  85. 'backbone.layer2.0.downsample.0.weight'
  86. 'backbone.layer2.0.downsample.1.weight'
  87. 'backbone.layer2.0.downsample.1.bias'
  88. 'backbone.layer2.0.downsample.1.running_mean'
  89. 'backbone.layer2.0.downsample.1.running_var'
  90. 'backbone.layer2.0.downsample.1.num_batches_tracked'
  91. 'backbone.layer2.1.conv1.weight'
  92. 'backbone.layer2.1.bn1.weight'
  93. 'backbone.layer2.1.bn1.bias'
  94. 'backbone.layer2.1.bn1.running_mean'
  95. 'backbone.layer2.1.bn1.running_var'
  96. 'backbone.layer2.1.bn1.num_batches_tracked'
  97. 'backbone.layer2.1.conv2.weight'
  98. 'backbone.layer2.1.bn2.weight'
  99. 'backbone.layer2.1.bn2.bias'
  100. 'backbone.layer2.1.bn2.running_mean'
  101. 'backbone.layer2.1.bn2.running_var'
  102. 'backbone.layer2.1.bn2.num_batches_tracked'
  103. 'backbone.layer2.1.conv3.weight'
  104. 'backbone.layer2.1.bn3.weight'
  105. 'backbone.layer2.1.bn3.bias'
  106. 'backbone.layer2.1.bn3.running_mean'
  107. 'backbone.layer2.1.bn3.running_var'
  108. 'backbone.layer2.1.bn3.num_batches_tracked'
  109. 'backbone.layer2.2.conv1.weight'
  110. 'backbone.layer2.2.bn1.weight'
  111. 'backbone.layer2.2.bn1.bias'
  112. 'backbone.layer2.2.bn1.running_mean'
  113. 'backbone.layer2.2.bn1.running_var'
  114. 'backbone.layer2.2.bn1.num_batches_tracked'
  115. 'backbone.layer2.2.conv2.weight'
  116. 'backbone.layer2.2.bn2.weight'
  117. 'backbone.layer2.2.bn2.bias'
  118. 'backbone.layer2.2.bn2.running_mean'
  119. 'backbone.layer2.2.bn2.running_var'
  120. 'backbone.layer2.2.bn2.num_batches_tracked'
  121. 'backbone.layer2.2.conv3.weight'
  122. 'backbone.layer2.2.bn3.weight'
  123. 'backbone.layer2.2.bn3.bias'
  124. 'backbone.layer2.2.bn3.running_mean'
  125. 'backbone.layer2.2.bn3.running_var'
  126. 'backbone.layer2.2.bn3.num_batches_tracked'
  127. 'backbone.layer2.3.conv1.weight'
  128. 'backbone.layer2.3.bn1.weight'
  129. 'backbone.layer2.3.bn1.bias'
  130. 'backbone.layer2.3.bn1.running_mean'
  131. 'backbone.layer2.3.bn1.running_var'
  132. 'backbone.layer2.3.bn1.num_batches_tracked'
  133. 'backbone.layer2.3.conv2.weight'
  134. 'backbone.layer2.3.bn2.weight'
  135. 'backbone.layer2.3.bn2.bias'
  136. 'backbone.layer2.3.bn2.running_mean'
  137. 'backbone.layer2.3.bn2.running_var'
  138. 'backbone.layer2.3.bn2.num_batches_tracked'
  139. 'backbone.layer2.3.conv3.weight'
  140. 'backbone.layer2.3.bn3.weight'
  141. 'backbone.layer2.3.bn3.bias'
  142. 'backbone.layer2.3.bn3.running_mean'
  143. 'backbone.layer2.3.bn3.running_var'
  144. 'backbone.layer2.3.bn3.num_batches_tracked'
  145. 'backbone.layer3.0.conv1.weight'
  146. 'backbone.layer3.0.bn1.weight'
  147. 'backbone.layer3.0.bn1.bias'
  148. 'backbone.layer3.0.bn1.running_mean'
  149. 'backbone.layer3.0.bn1.running_var'
  150. 'backbone.layer3.0.bn1.num_batches_tracked'
  151. 'backbone.layer3.0.conv2.weight'
  152. 'backbone.layer3.0.bn2.weight'
  153. 'backbone.layer3.0.bn2.bias'
  154. 'backbone.layer3.0.bn2.running_mean'
  155. 'backbone.layer3.0.bn2.running_var'
  156. 'backbone.layer3.0.bn2.num_batches_tracked'
  157. 'backbone.layer3.0.conv3.weight'
  158. 'backbone.layer3.0.bn3.weight'
  159. 'backbone.layer3.0.bn3.bias'
  160. 'backbone.layer3.0.bn3.running_mean'
  161. 'backbone.layer3.0.bn3.running_var'
  162. 'backbone.layer3.0.bn3.num_batches_tracked'
  163. 'backbone.layer3.0.downsample.0.weight'
  164. 'backbone.layer3.0.downsample.1.weight'
  165. 'backbone.layer3.0.downsample.1.bias'
  166. 'backbone.layer3.0.downsample.1.running_mean'
  167. 'backbone.layer3.0.downsample.1.running_var'
  168. 'backbone.layer3.0.downsample.1.num_batches_tracked'
  169. 'backbone.layer3.1.conv1.weight'
  170. 'backbone.layer3.1.bn1.weight'
  171. 'backbone.layer3.1.bn1.bias'
  172. 'backbone.layer3.1.bn1.running_mean'
  173. 'backbone.layer3.1.bn1.running_var'
  174. 'backbone.layer3.1.bn1.num_batches_tracked'
  175. 'backbone.layer3.1.conv2.weight'
  176. 'backbone.layer3.1.bn2.weight'
  177. 'backbone.layer3.1.bn2.bias'
  178. 'backbone.layer3.1.bn2.running_mean'
  179. 'backbone.layer3.1.bn2.running_var'
  180. 'backbone.layer3.1.bn2.num_batches_tracked'
  181. 'backbone.layer3.1.conv3.weight'
  182. 'backbone.layer3.1.bn3.weight'
  183. 'backbone.layer3.1.bn3.bias'
  184. 'backbone.layer3.1.bn3.running_mean'
  185. 'backbone.layer3.1.bn3.running_var'
  186. 'backbone.layer3.1.bn3.num_batches_tracked'
  187. 'backbone.layer3.2.conv1.weight'
  188. 'backbone.layer3.2.bn1.weight'
  189. 'backbone.layer3.2.bn1.bias'
  190. 'backbone.layer3.2.bn1.running_mean'
  191. 'backbone.layer3.2.bn1.running_var'
  192. 'backbone.layer3.2.bn1.num_batches_tracked'
  193. 'backbone.layer3.2.conv2.weight'
  194. 'backbone.layer3.2.bn2.weight'
  195. 'backbone.layer3.2.bn2.bias'
  196. 'backbone.layer3.2.bn2.running_mean'
  197. 'backbone.layer3.2.bn2.running_var'
  198. 'backbone.layer3.2.bn2.num_batches_tracked'
  199. 'backbone.layer3.2.conv3.weight'
  200. 'backbone.layer3.2.bn3.weight'
  201. 'backbone.layer3.2.bn3.bias'
  202. 'backbone.layer3.2.bn3.running_mean'
  203. 'backbone.layer3.2.bn3.running_var'
  204. 'backbone.layer3.2.bn3.num_batches_tracked'
  205. 'backbone.layer3.3.conv1.weight'
  206. 'backbone.layer3.3.bn1.weight'
  207. 'backbone.layer3.3.bn1.bias'
  208. 'backbone.layer3.3.bn1.running_mean'
  209. 'backbone.layer3.3.bn1.running_var'
  210. 'backbone.layer3.3.bn1.num_batches_tracked'
  211. 'backbone.layer3.3.conv2.weight'
  212. 'backbone.layer3.3.bn2.weight'
  213. 'backbone.layer3.3.bn2.bias'
  214. 'backbone.layer3.3.bn2.running_mean'
  215. 'backbone.layer3.3.bn2.running_var'
  216. 'backbone.layer3.3.bn2.num_batches_tracked'
  217. 'backbone.layer3.3.conv3.weight'
  218. 'backbone.layer3.3.bn3.weight'
  219. 'backbone.layer3.3.bn3.bias'
  220. 'backbone.layer3.3.bn3.running_mean'
  221. 'backbone.layer3.3.bn3.running_var'
  222. 'backbone.layer3.3.bn3.num_batches_tracked'
  223. 'backbone.layer3.4.conv1.weight'
  224. 'backbone.layer3.4.bn1.weight'
  225. 'backbone.layer3.4.bn1.bias'
  226. 'backbone.layer3.4.bn1.running_mean'
  227. 'backbone.layer3.4.bn1.running_var'
  228. 'backbone.layer3.4.bn1.num_batches_tracked'
  229. 'backbone.layer3.4.conv2.weight'
  230. 'backbone.layer3.4.bn2.weight'
  231. 'backbone.layer3.4.bn2.bias'
  232. 'backbone.layer3.4.bn2.running_mean'
  233. 'backbone.layer3.4.bn2.running_var'
  234. 'backbone.layer3.4.bn2.num_batches_tracked'
  235. 'backbone.layer3.4.conv3.weight'
  236. 'backbone.layer3.4.bn3.weight'
  237. 'backbone.layer3.4.bn3.bias'
  238. 'backbone.layer3.4.bn3.running_mean'
  239. 'backbone.layer3.4.bn3.running_var'
  240. 'backbone.layer3.4.bn3.num_batches_tracked'
  241. 'backbone.layer3.5.conv1.weight'
  242. 'backbone.layer3.5.bn1.weight'
  243. 'backbone.layer3.5.bn1.bias'
  244. 'backbone.layer3.5.bn1.running_mean'
  245. 'backbone.layer3.5.bn1.running_var'
  246. 'backbone.layer3.5.bn1.num_batches_tracked'
  247. 'backbone.layer3.5.conv2.weight'
  248. 'backbone.layer3.5.bn2.weight'
  249. 'backbone.layer3.5.bn2.bias'
  250. 'backbone.layer3.5.bn2.running_mean'
  251. 'backbone.layer3.5.bn2.running_var'
  252. 'backbone.layer3.5.bn2.num_batches_tracked'
  253. 'backbone.layer3.5.conv3.weight'
  254. 'backbone.layer3.5.bn3.weight'
  255. 'backbone.layer3.5.bn3.bias'
  256. 'backbone.layer3.5.bn3.running_mean'
  257. 'backbone.layer3.5.bn3.running_var'
  258. 'backbone.layer3.5.bn3.num_batches_tracked'
  259. 'backbone.layer4.0.conv1.weight'
  260. 'backbone.layer4.0.bn1.weight'
  261. 'backbone.layer4.0.bn1.bias'
  262. 'backbone.layer4.0.bn1.running_mean'
  263. 'backbone.layer4.0.bn1.running_var'
  264. 'backbone.layer4.0.bn1.num_batches_tracked'
  265. 'backbone.layer4.0.conv2.weight'
  266. 'backbone.layer4.0.bn2.weight'
  267. 'backbone.layer4.0.bn2.bias'
  268. 'backbone.layer4.0.bn2.running_mean'
  269. 'backbone.layer4.0.bn2.running_var'
  270. 'backbone.layer4.0.bn2.num_batches_tracked'
  271. 'backbone.layer4.0.conv3.weight'
  272. 'backbone.layer4.0.bn3.weight'
  273. 'backbone.layer4.0.bn3.bias'
  274. 'backbone.layer4.0.bn3.running_mean'
  275. 'backbone.layer4.0.bn3.running_var'
  276. 'backbone.layer4.0.bn3.num_batches_tracked'
  277. 'backbone.layer4.0.downsample.0.weight'
  278. 'backbone.layer4.0.downsample.1.weight'
  279. 'backbone.layer4.0.downsample.1.bias'
  280. 'backbone.layer4.0.downsample.1.running_mean'
  281. 'backbone.layer4.0.downsample.1.running_var'
  282. 'backbone.layer4.0.downsample.1.num_batches_tracked'
  283. 'backbone.layer4.1.conv1.weight'
  284. 'backbone.layer4.1.bn1.weight'
  285. 'backbone.layer4.1.bn1.bias'
  286. 'backbone.layer4.1.bn1.running_mean'
  287. 'backbone.layer4.1.bn1.running_var'
  288. 'backbone.layer4.1.bn1.num_batches_tracked'
  289. 'backbone.layer4.1.conv2.weight'
  290. 'backbone.layer4.1.bn2.weight'
  291. 'backbone.layer4.1.bn2.bias'
  292. 'backbone.layer4.1.bn2.running_mean'
  293. 'backbone.layer4.1.bn2.running_var'
  294. 'backbone.layer4.1.bn2.num_batches_tracked'
  295. 'backbone.layer4.1.conv3.weight'
  296. 'backbone.layer4.1.bn3.weight'
  297. 'backbone.layer4.1.bn3.bias'
  298. 'backbone.layer4.1.bn3.running_mean'
  299. 'backbone.layer4.1.bn3.running_var'
  300. 'backbone.layer4.1.bn3.num_batches_tracked'
  301. 'backbone.layer4.2.conv1.weight'
  302. 'backbone.layer4.2.bn1.weight'
  303. 'backbone.layer4.2.bn1.bias'
  304. 'backbone.layer4.2.bn1.running_mean'
  305. 'backbone.layer4.2.bn1.running_var'
  306. 'backbone.layer4.2.bn1.num_batches_tracked'
  307. 'backbone.layer4.2.conv2.weight'
  308. 'backbone.layer4.2.bn2.weight'
  309. 'backbone.layer4.2.bn2.bias'
  310. 'backbone.layer4.2.bn2.running_mean'
  311. 'backbone.layer4.2.bn2.running_var'
  312. 'backbone.layer4.2.bn2.num_batches_tracked'
  313. 'backbone.layer4.2.conv3.weight'
  314. 'backbone.layer4.2.bn3.weight'
  315. 'backbone.layer4.2.bn3.bias'
  316. 'backbone.layer4.2.bn3.running_mean'
  317. 'backbone.layer4.2.bn3.running_var'
  318. 'backbone.layer4.2.bn3.num_batches_tracked'
  319. 'neck.lateral_convs.0.conv.weight'
  320. 'neck.lateral_convs.0.conv.bias'
  321. 'neck.lateral_convs.1.conv.weight'
  322. 'neck.lateral_convs.1.conv.bias'
  323. 'neck.lateral_convs.2.conv.weight'
  324. 'neck.lateral_convs.2.conv.bias'
  325. 'neck.lateral_convs.3.conv.weight'
  326. 'neck.lateral_convs.3.conv.bias'
  327. 'neck.fpn_convs.0.conv.weight'
  328. 'neck.fpn_convs.0.conv.bias'
  329. 'neck.fpn_convs.1.conv.weight'
  330. 'neck.fpn_convs.1.conv.bias'
  331. 'neck.fpn_convs.2.conv.weight'
  332. 'neck.fpn_convs.2.conv.bias'
  333. 'neck.fpn_convs.3.conv.weight'
  334. 'neck.fpn_convs.3.conv.bias'
  335. 'rpn_head.rpn_conv.weight'
  336. 'rpn_head.rpn_conv.bias'
  337. 'rpn_head.rpn_cls.weight'
  338. 'rpn_head.rpn_cls.bias'
  339. 'rpn_head.rpn_reg.weight'
  340. 'rpn_head.rpn_reg.bias'
  341. 'roi_head.bbox_head.fc_cls.weight'
  342. 'roi_head.bbox_head.fc_cls.bias'
  343. 'roi_head.bbox_head.fc_reg.weight'
  344. 'roi_head.bbox_head.fc_reg.bias'
  345. 'roi_head.bbox_head.shared_fcs.0.weight'
  346. 'roi_head.bbox_head.shared_fcs.0.bias'
  347. 'roi_head.bbox_head.shared_fcs.1.weight'
  348. 'roi_head.bbox_head.shared_fcs.1.bias'

Second List (keys of state_dict from newly initialized model ):

  1. 'detector.backbone.conv1.weight',
  2. 'detector.backbone.bn1.weight',
  3. 'detector.backbone.bn1.bias',
  4. 'detector.backbone.bn1.running_mean',
  5. 'detector.backbone.bn1.running_var',
  6. 'detector.backbone.bn1.num_batches_tracked',
  7. 'detector.backbone.layer1.0.conv1.weight',
  8. 'detector.backbone.layer1.0.bn1.weight',
  9. 'detector.backbone.layer1.0.bn1.bias',
  10. 'detector.backbone.layer1.0.bn1.running_mean',
  11. 'detector.backbone.layer1.0.bn1.running_var',
  12. 'detector.backbone.layer1.0.bn1.num_batches_tracked',
  13. 'detector.backbone.layer1.0.conv2.weight',
  14. 'detector.backbone.layer1.0.bn2.weight',
  15. 'detector.backbone.layer1.0.bn2.bias',
  16. 'detector.backbone.layer1.0.bn2.running_mean',
  17. 'detector.backbone.layer1.0.bn2.running_var',
  18. 'detector.backbone.layer1.0.bn2.num_batches_tracked',
  19. 'detector.backbone.layer1.0.conv3.weight',
  20. 'detector.backbone.layer1.0.bn3.weight',
  21. 'detector.backbone.layer1.0.bn3.bias',
  22. 'detector.backbone.layer1.0.bn3.running_mean',
  23. 'detector.backbone.layer1.0.bn3.running_var',
  24. 'detector.backbone.layer1.0.bn3.num_batches_tracked',
  25. 'detector.backbone.layer1.0.downsample.0.weight',
  26. 'detector.backbone.layer1.0.downsample.1.weight',
  27. 'detector.backbone.layer1.0.downsample.1.bias',
  28. 'detector.backbone.layer1.0.downsample.1.running_mean',
  29. 'detector.backbone.layer1.0.downsample.1.running_var',
  30. 'detector.backbone.layer1.0.downsample.1.num_batches_tracked',
  31. 'detector.backbone.layer1.1.conv1.weight',
  32. 'detector.backbone.layer1.1.bn1.weight',
  33. 'detector.backbone.layer1.1.bn1.bias',
  34. 'detector.backbone.layer1.1.bn1.running_mean',
  35. 'detector.backbone.layer1.1.bn1.running_var',
  36. 'detector.backbone.layer1.1.bn1.num_batches_tracked',
  37. 'detector.backbone.layer1.1.conv2.weight',
  38. 'detector.backbone.layer1.1.bn2.weight',
  39. 'detector.backbone.layer1.1.bn2.bias',
  40. 'detector.backbone.layer1.1.bn2.running_mean',
  41. 'detector.backbone.layer1.1.bn2.running_var',
  42. 'detector.backbone.layer1.1.bn2.num_batches_tracked',
  43. 'detector.backbone.layer1.1.conv3.weight',
  44. 'detector.backbone.layer1.1.bn3.weight',
  45. 'detector.backbone.layer1.1.bn3.bias',
  46. 'detector.backbone.layer1.1.bn3.running_mean',
  47. 'detector.backbone.layer1.1.bn3.running_var',
  48. 'detector.backbone.layer1.1.bn3.num_batches_tracked',
  49. 'detector.backbone.layer1.2.conv1.weight',
  50. 'detector.backbone.layer1.2.bn1.weight',
  51. 'detector.backbone.layer1.2.bn1.bias',
  52. 'detector.backbone.layer1.2.bn1.running_mean',
  53. 'detector.backbone.layer1.2.bn1.running_var',
  54. 'detector.backbone.layer1.2.bn1.num_batches_tracked',
  55. 'detector.backbone.layer1.2.conv2.weight',
  56. 'detector.backbone.layer1.2.bn2.weight',
  57. 'detector.backbone.layer1.2.bn2.bias',
  58. 'detector.backbone.layer1.2.bn2.running_mean',
  59. 'detector.backbone.layer1.2.bn2.running_var',
  60. 'detector.backbone.layer1.2.bn2.num_batches_tracked',
  61. 'detector.backbone.layer1.2.conv3.weight',
  62. 'detector.backbone.layer1.2.bn3.weight',
  63. 'detector.backbone.layer1.2.bn3.bias',
  64. 'detector.backbone.layer1.2.bn3.running_mean',
  65. 'detector.backbone.layer1.2.bn3.running_var',
  66. 'detector.backbone.layer1.2.bn3.num_batches_tracked',
  67. 'detector.backbone.layer2.0.conv1.weight',
  68. 'detector.backbone.layer2.0.bn1.weight',
  69. 'detector.backbone.layer2.0.bn1.bias',
  70. 'detector.backbone.layer2.0.bn1.running_mean',
  71. 'detector.backbone.layer2.0.bn1.running_var',
  72. 'detector.backbone.layer2.0.bn1.num_batches_tracked',
  73. 'detector.backbone.layer2.0.conv2.weight',
  74. 'detector.backbone.layer2.0.bn2.weight',
  75. 'detector.backbone.layer2.0.bn2.bias',
  76. 'detector.backbone.layer2.0.bn2.running_mean',
  77. 'detector.backbone.layer2.0.bn2.running_var',
  78. 'detector.backbone.layer2.0.bn2.num_batches_tracked',
  79. 'detector.backbone.layer2.0.conv3.weight',
  80. 'detector.backbone.layer2.0.bn3.weight',
  81. 'detector.backbone.layer2.0.bn3.bias',
  82. 'detector.backbone.layer2.0.bn3.running_mean',
  83. 'detector.backbone.layer2.0.bn3.running_var',
  84. 'detector.backbone.layer2.0.bn3.num_batches_tracked',
  85. 'detector.backbone.layer2.0.downsample.0.weight',
  86. 'detector.backbone.layer2.0.downsample.1.weight',
  87. 'detector.backbone.layer2.0.downsample.1.bias',
  88. 'detector.backbone.layer2.0.downsample.1.running_mean',
  89. 'detector.backbone.layer2.0.downsample.1.running_var',
  90. 'detector.backbone.layer2.0.downsample.1.num_batches_tracked',
  91. 'detector.backbone.layer2.1.conv1.weight',
  92. 'detector.backbone.layer2.1.bn1.weight',
  93. 'detector.backbone.layer2.1.bn1.bias',
  94. 'detector.backbone.layer2.1.bn1.running_mean',
  95. 'detector.backbone.layer2.1.bn1.running_var',
  96. 'detector.backbone.layer2.1.bn1.num_batches_tracked',
  97. 'detector.backbone.layer2.1.conv2.weight',
  98. 'detector.backbone.layer2.1.bn2.weight',
  99. 'detector.backbone.layer2.1.bn2.bias',
  100. 'detector.backbone.layer2.1.bn2.running_mean',
  101. 'detector.backbone.layer2.1.bn2.running_var',
  102. 'detector.backbone.layer2.1.bn2.num_batches_tracked',
  103. 'detector.backbone.layer2.1.conv3.weight',
  104. 'detector.backbone.layer2.1.bn3.weight',
  105. 'detector.backbone.layer2.1.bn3.bias',
  106. 'detector.backbone.layer2.1.bn3.running_mean',
  107. 'detector.backbone.layer2.1.bn3.running_var',
  108. 'detector.backbone.layer2.1.bn3.num_batches_tracked',
  109. 'detector.backbone.layer2.2.conv1.weight',
  110. 'detector.backbone.layer2.2.bn1.weight',
  111. 'detector.backbone.layer2.2.bn1.bias',
  112. 'detector.backbone.layer2.2.bn1.running_mean',
  113. 'detector.backbone.layer2.2.bn1.running_var',
  114. 'detector.backbone.layer2.2.bn1.num_batches_tracked',
  115. 'detector.backbone.layer2.2.conv2.weight',
  116. 'detector.backbone.layer2.2.bn2.weight',
  117. 'detector.backbone.layer2.2.bn2.bias',
  118. 'detector.backbone.layer2.2.bn2.running_mean',
  119. 'detector.backbone.layer2.2.bn2.running_var',
  120. 'detector.backbone.layer2.2.bn2.num_batches_tracked',
  121. 'detector.backbone.layer2.2.conv3.weight',
  122. 'detector.backbone.layer2.2.bn3.weight',
  123. 'detector.backbone.layer2.2.bn3.bias',
  124. 'detector.backbone.layer2.2.bn3.running_mean',
  125. 'detector.backbone.layer2.2.bn3.running_var',
  126. 'detector.backbone.layer2.2.bn3.num_batches_tracked',
  127. 'detector.backbone.layer2.3.conv1.weight',
  128. 'detector.backbone.layer2.3.bn1.weight',
  129. 'detector.backbone.layer2.3.bn1.bias',
  130. 'detector.backbone.layer2.3.bn1.running_mean',
  131. 'detector.backbone.layer2.3.bn1.running_var',
  132. 'detector.backbone.layer2.3.bn1.num_batches_tracked',
  133. 'detector.backbone.layer2.3.conv2.weight',
  134. 'detector.backbone.layer2.3.bn2.weight',
  135. 'detector.backbone.layer2.3.bn2.bias',
  136. 'detector.backbone.layer2.3.bn2.running_mean',
  137. 'detector.backbone.layer2.3.bn2.running_var',
  138. 'detector.backbone.layer2.3.bn2.num_batches_tracked',
  139. 'detector.backbone.layer2.3.conv3.weight',
  140. 'detector.backbone.layer2.3.bn3.weight',
  141. 'detector.backbone.layer2.3.bn3.bias',
  142. 'detector.backbone.layer2.3.bn3.running_mean',
  143. 'detector.backbone.layer2.3.bn3.running_var',
  144. 'detector.backbone.layer2.3.bn3.num_batches_tracked',
  145. 'detector.backbone.layer3.0.conv1.weight',
  146. 'detector.backbone.layer3.0.bn1.weight',
  147. 'detector.backbone.layer3.0.bn1.bias',
  148. 'detector.backbone.layer3.0.bn1.running_mean',
  149. 'detector.backbone.layer3.0.bn1.running_var',
  150. 'detector.backbone.layer3.0.bn1.num_batches_tracked',
  151. 'detector.backbone.layer3.0.conv2.weight',
  152. 'detector.backbone.layer3.0.bn2.weight',
  153. 'detector.backbone.layer3.0.bn2.bias',
  154. 'detector.backbone.layer3.0.bn2.running_mean',
  155. 'detector.backbone.layer3.0.bn2.running_var',
  156. 'detector.backbone.layer3.0.bn2.num_batches_tracked',
  157. 'detector.backbone.layer3.0.conv3.weight',
  158. 'detector.backbone.layer3.0.bn3.weight',
  159. 'detector.backbone.layer3.0.bn3.bias',
  160. 'detector.backbone.layer3.0.bn3.running_mean',
  161. 'detector.backbone.layer3.0.bn3.running_var',
  162. 'detector.backbone.layer3.0.bn3.num_batches_tracked',
  163. 'detector.backbone.layer3.0.downsample.0.weight',
  164. 'detector.backbone.layer3.0.downsample.1.weight',
  165. 'detector.backbone.layer3.0.downsample.1.bias',
  166. 'detector.backbone.layer3.0.downsample.1.running_mean',
  167. 'detector.backbone.layer3.0.downsample.1.running_var',
  168. 'detector.backbone.layer3.0.downsample.1.num_batches_tracked',
  169. 'detector.backbone.layer3.1.conv1.weight',
  170. 'detector.backbone.layer3.1.bn1.weight',
  171. 'detector.backbone.layer3.1.bn1.bias',
  172. 'detector.backbone.layer3.1.bn1.running_mean',
  173. 'detector.backbone.layer3.1.bn1.running_var',
  174. 'detector.backbone.layer3.1.bn1.num_batches_tracked',
  175. 'detector.backbone.layer3.1.conv2.weight',
  176. 'detector.backbone.layer3.1.bn2.weight',
  177. 'detector.backbone.layer3.1.bn2.bias',
  178. 'detector.backbone.layer3.1.bn2.running_mean',
  179. 'detector.backbone.layer3.1.bn2.running_var',
  180. 'detector.backbone.layer3.1.bn2.num_batches_tracked',
  181. 'detector.backbone.layer3.1.conv3.weight',
  182. 'detector.backbone.layer3.1.bn3.weight',
  183. 'detector.backbone.layer3.1.bn3.bias',
  184. 'detector.backbone.layer3.1.bn3.running_mean',
  185. 'detector.backbone.layer3.1.bn3.running_var',
  186. 'detector.backbone.layer3.1.bn3.num_batches_tracked',
  187. 'detector.backbone.layer3.2.conv1.weight',
  188. 'detector.backbone.layer3.2.bn1.weight',
  189. 'detector.backbone.layer3.2.bn1.bias',
  190. 'detector.backbone.layer3.2.bn1.running_mean',
  191. 'detector.backbone.layer3.2.bn1.running_var',
  192. 'detector.backbone.layer3.2.bn1.num_batches_tracked',
  193. 'detector.backbone.layer3.2.conv2.weight',
  194. 'detector.backbone.layer3.2.bn2.weight',
  195. 'detector.backbone.layer3.2.bn2.bias',
  196. 'detector.backbone.layer3.2.bn2.running_mean',
  197. 'detector.backbone.layer3.2.bn2.running_var',
  198. 'detector.backbone.layer3.2.bn2.num_batches_tracked',
  199. 'detector.backbone.layer3.2.conv3.weight',
  200. 'detector.backbone.layer3.2.bn3.weight',
  201. 'detector.backbone.layer3.2.bn3.bias',
  202. 'detector.backbone.layer3.2.bn3.running_mean',
  203. 'detector.backbone.layer3.2.bn3.running_var',
  204. 'detector.backbone.layer3.2.bn3.num_batches_tracked',
  205. 'detector.backbone.layer3.3.conv1.weight',
  206. 'detector.backbone.layer3.3.bn1.weight',
  207. 'detector.backbone.layer3.3.bn1.bias',
  208. 'detector.backbone.layer3.3.bn1.running_mean',
  209. 'detector.backbone.layer3.3.bn1.running_var',
  210. 'detector.backbone.layer3.3.bn1.num_batches_tracked',
  211. 'detector.backbone.layer3.3.conv2.weight',
  212. 'detector.backbone.layer3.3.bn2.weight',
  213. 'detector.backbone.layer3.3.bn2.bias',
  214. 'detector.backbone.layer3.3.bn2.running_mean',
  215. 'detector.backbone.layer3.3.bn2.running_var',
  216. 'detector.backbone.layer3.3.bn2.num_batches_tracked',
  217. 'detector.backbone.layer3.3.conv3.weight',
  218. 'detector.backbone.layer3.3.bn3.weight',
  219. 'detector.backbone.layer3.3.bn3.bias',
  220. 'detector.backbone.layer3.3.bn3.running_mean',
  221. 'detector.backbone.layer3.3.bn3.running_var',
  222. 'detector.backbone.layer3.3.bn3.num_batches_tracked',
  223. 'detector.backbone.layer3.4.conv1.weight',
  224. 'detector.backbone.layer3.4.bn1.weight',
  225. 'detector.backbone.layer3.4.bn1.bias',
  226. 'detector.backbone.layer3.4.bn1.running_mean',
  227. 'detector.backbone.layer3.4.bn1.running_var',
  228. 'detector.backbone.layer3.4.bn1.num_batches_tracked',
  229. 'detector.backbone.layer3.4.conv2.weight',
  230. 'detector.backbone.layer3.4.bn2.weight',
  231. 'detector.backbone.layer3.4.bn2.bias',
  232. 'detector.backbone.layer3.4.bn2.running_mean',
  233. 'detector.backbone.layer3.4.bn2.running_var',
  234. 'detector.backbone.layer3.4.bn2.num_batches_tracked',
  235. 'detector.backbone.layer3.4.conv3.weight',
  236. 'detector.backbone.layer3.4.bn3.weight',
  237. 'detector.backbone.layer3.4.bn3.bias',
  238. 'detector.backbone.layer3.4.bn3.running_mean',
  239. 'detector.backbone.layer3.4.bn3.running_var',
  240. 'detector.backbone.layer3.4.bn3.num_batches_tracked',
  241. 'detector.backbone.layer3.5.conv1.weight',
  242. 'detector.backbone.layer3.5.bn1.weight',
  243. 'detector.backbone.layer3.5.bn1.bias',
  244. 'detector.backbone.layer3.5.bn1.running_mean',
  245. 'detector.backbone.layer3.5.bn1.running_var',
  246. 'detector.backbone.layer3.5.bn1.num_batches_tracked',
  247. 'detector.backbone.layer3.5.conv2.weight',
  248. 'detector.backbone.layer3.5.bn2.weight',
  249. 'detector.backbone.layer3.5.bn2.bias',
  250. 'detector.backbone.layer3.5.bn2.running_mean',
  251. 'detector.backbone.layer3.5.bn2.running_var',
  252. 'detector.backbone.layer3.5.bn2.num_batches_tracked',
  253. 'detector.backbone.layer3.5.conv3.weight',
  254. 'detector.backbone.layer3.5.bn3.weight',
  255. 'detector.backbone.layer3.5.bn3.bias',
  256. 'detector.backbone.layer3.5.bn3.running_mean',
  257. 'detector.backbone.layer3.5.bn3.running_var',
  258. 'detector.backbone.layer3.5.bn3.num_batches_tracked',
  259. 'detector.backbone.layer4.0.conv1.weight',
  260. 'detector.backbone.layer4.0.bn1.weight',
  261. 'detector.backbone.layer4.0.bn1.bias',
  262. 'detector.backbone.layer4.0.bn1.running_mean',
  263. 'detector.backbone.layer4.0.bn1.running_var',
  264. 'detector.backbone.layer4.0.bn1.num_batches_tracked',
  265. 'detector.backbone.layer4.0.conv2.weight',
  266. 'detector.backbone.layer4.0.bn2.weight',
  267. 'detector.backbone.layer4.0.bn2.bias',
  268. 'detector.backbone.layer4.0.bn2.running_mean',
  269. 'detector.backbone.layer4.0.bn2.running_var',
  270. 'detector.backbone.layer4.0.bn2.num_batches_tracked',
  271. 'detector.backbone.layer4.0.conv3.weight',
  272. 'detector.backbone.layer4.0.bn3.weight',
  273. 'detector.backbone.layer4.0.bn3.bias',
  274. 'detector.backbone.layer4.0.bn3.running_mean',
  275. 'detector.backbone.layer4.0.bn3.running_var',
  276. 'detector.backbone.layer4.0.bn3.num_batches_tracked',
  277. 'detector.backbone.layer4.0.downsample.0.weight',
  278. 'detector.backbone.layer4.0.downsample.1.weight',
  279. 'detector.backbone.layer4.0.downsample.1.bias',
  280. 'detector.backbone.layer4.0.downsample.1.running_mean',
  281. 'detector.backbone.layer4.0.downsample.1.running_var',
  282. 'detector.backbone.layer4.0.downsample.1.num_batches_tracked',
  283. 'detector.backbone.layer4.1.conv1.weight',
  284. 'detector.backbone.layer4.1.bn1.weight',
  285. 'detector.backbone.layer4.1.bn1.bias',
  286. 'detector.backbone.layer4.1.bn1.running_mean',
  287. 'detector.backbone.layer4.1.bn1.running_var',
  288. 'detector.backbone.layer4.1.bn1.num_batches_tracked',
  289. 'detector.backbone.layer4.1.conv2.weight',
  290. 'detector.backbone.layer4.1.bn2.weight',
  291. 'detector.backbone.layer4.1.bn2.bias',
  292. 'detector.backbone.layer4.1.bn2.running_mean',
  293. 'detector.backbone.layer4.1.bn2.running_var',
  294. 'detector.backbone.layer4.1.bn2.num_batches_tracked',
  295. 'detector.backbone.layer4.1.conv3.weight',
  296. 'detector.backbone.layer4.1.bn3.weight',
  297. 'detector.backbone.layer4.1.bn3.bias',
  298. 'detector.backbone.layer4.1.bn3.running_mean',
  299. 'detector.backbone.layer4.1.bn3.running_var',
  300. 'detector.backbone.layer4.1.bn3.num_batches_tracked',
  301. 'detector.backbone.layer4.2.conv1.weight',
  302. 'detector.backbone.layer4.2.bn1.weight',
  303. 'detector.backbone.layer4.2.bn1.bias',
  304. 'detector.backbone.layer4.2.bn1.running_mean',
  305. 'detector.backbone.layer4.2.bn1.running_var',
  306. 'detector.backbone.layer4.2.bn1.num_batches_tracked',
  307. 'detector.backbone.layer4.2.conv2.weight',
  308. 'detector.backbone.layer4.2.bn2.weight',
  309. 'detector.backbone.layer4.2.bn2.bias',
  310. 'detector.backbone.layer4.2.bn2.running_mean',
  311. 'detector.backbone.layer4.2.bn2.running_var',
  312. 'detector.backbone.layer4.2.bn2.num_batches_tracked',
  313. 'detector.backbone.layer4.2.conv3.weight',
  314. 'detector.backbone.layer4.2.bn3.weight',
  315. 'detector.backbone.layer4.2.bn3.bias',
  316. 'detector.backbone.layer4.2.bn3.running_mean',
  317. 'detector.backbone.layer4.2.bn3.running_var',
  318. 'detector.backbone.layer4.2.bn3.num_batches_tracked',
  319. 'detector.neck.lateral_convs.0.conv.weight',
  320. 'detector.neck.lateral_convs.0.conv.bias',
  321. 'detector.neck.lateral_convs.1.conv.weight',
  322. 'detector.neck.lateral_convs.1.conv.bias',
  323. 'detector.neck.lateral_convs.2.conv.weight',
  324. 'detector.neck.lateral_convs.2.conv.bias',
  325. 'detector.neck.lateral_convs.3.conv.weight',
  326. 'detector.neck.lateral_convs.3.conv.bias',
  327. 'detector.neck.fpn_convs.0.conv.weight',
  328. 'detector.neck.fpn_convs.0.conv.bias',
  329. 'detector.neck.fpn_convs.1.conv.weight',
  330. 'detector.neck.fpn_convs.1.conv.bias',
  331. 'detector.neck.fpn_convs.2.conv.weight',
  332. 'detector.neck.fpn_convs.2.conv.bias',
  333. 'detector.neck.fpn_convs.3.conv.weight',
  334. 'detector.neck.fpn_convs.3.conv.bias',
  335. 'detector.rpn_head.rpn_conv.weight',
  336. 'detector.rpn_head.rpn_conv.bias',
  337. 'detector.rpn_head.rpn_cls.weight',
  338. 'detector.rpn_head.rpn_cls.bias',
  339. 'detector.rpn_head.rpn_reg.weight',
  340. 'detector.rpn_head.rpn_reg.bias',
  341. 'detector.roi_head.bbox_head.fc_cls.weight',
  342. 'detector.roi_head.bbox_head.fc_cls.bias',
  343. 'detector.roi_head.bbox_head.fc_reg.weight',
  344. 'detector.roi_head.bbox_head.fc_reg.bias',
  345. 'detector.roi_head.bbox_head.shared_fcs.0.weight',
  346. 'detector.roi_head.bbox_head.shared_fcs.0.bias',
  347. 'detector.roi_head.bbox_head.shared_fcs.1.weight',
  348. 'detector.roi_head.bbox_head.shared_fcs.1.bias',
  349. 'track_head.track_head.convs.0.conv.weight', # missing
  350. 'track_head.track_head.convs.0.gn.weight', # missing
  351. 'track_head.track_head.convs.0.gn.bias', # missing
  352. 'track_head.track_head.convs.1.conv.weight', # missing
  353. 'track_head.track_head.convs.1.gn.weight', # missing
  354. 'track_head.track_head.convs.1.gn.bias', # missing
  355. 'track_head.track_head.convs.2.conv.weight', # missing
  356. 'track_head.track_head.convs.2.gn.weight', # missing
  357. 'track_head.track_head.convs.2.gn.bias', # missing
  358. 'track_head.track_head.convs.3.conv.weight',# missing
  359. 'track_head.track_head.convs.3.gn.weight', # missing
  360. 'track_head.track_head.convs.3.gn.bias', # missing
  361. 'track_head.track_head.fcs.0.weight', # missing
  362. 'track_head.track_head.fcs.0.bias', # missing
  363. 'track_head.track_head.fc_embed.weight', # missing
  364. 'track_head.track_head.fc_embed.bias' # missing
Raspberry-beans commented 1 year ago

Hi, The following link contains weights file corresponding to given config files.

tanjary21 commented 1 year ago

Hi,

Thank you for the prompt response.

That checkpoint appears to be for the bdd100k-trained model. I'm actually looking for the MOT17-trained checkpoint. Do you think you can send me the link to that(with trained weights for the embedding head)?

Raspberry-beans commented 1 year ago

Hi, I actually don't know about MOT17 weights. Also in the repo we have weights only for BDD and Tao dataset. One way could be to train on MOT17 with given MOT17 config file.

Honestly I am still struggling to make all commands (train, test, inference, evaluate etc.) running properly.

tanjary21 commented 1 year ago

Hi,

I understand.

I've been using the older version of QDTrack from 2021. I've trained it myself but I'm really interested in the new checkpoints because it's pretrained on crowdhuman, which I've done myself as well but I can't seem to reproduce results.

In any case, yea, working with this repo is proving difficult if you want to do customized stuff, due to the extreme abstraction of MMCV.

Anyways, I'd appreciate any kind of help. No rush. I'm just interested in the MOT17 checkpoint that you use to get the new results.

On Thu, May 18, 2023 at 1:26 PM Raspberry-beans @.***> wrote:

Hi, I actually don't know about MOT17 weights. Also in the repo we have weights only for BDD and Tao dataset. One way could be to train on MOT17 with given MOT17 config file.

Honestly I am still struggling to make all commands (train, test, inference, evaluate etc.) running properly.

— Reply to this email directly, view it on GitHub https://github.com/SysCV/qdtrack/issues/138#issuecomment-1552851502, or unsubscribe https://github.com/notifications/unsubscribe-auth/AS4IMFWZIYLD7XLUYO2DKB3XGX2N5ANCNFSM6AAAAAAYE2NAZM . You are receiving this because you authored the thread.Message ID: @.***>

Raspberry-beans commented 1 year ago

Yeah mmcv extraction is a headache.

I have to focus on BDD dataset. Till now only my inference is successfully using qdtrack-frcnn_r50_fpn_12e_bdd100k.py config file and checkpoint from the provided link. I did not train it myself on BDD, just used the trained weights.

But now facing issues to reproduce the evaluation metrics result on BDD like mMOTA etc.

valeriocost commented 10 months ago

Yeah mmcv extraction is a headache.

I have to focus on BDD dataset. Till now only my inference is successfully using qdtrack-frcnn_r50_fpn_12e_bdd100k.py config file and checkpoint from the provided link. I did not train it myself on BDD, just used the trained weights.

But now facing issues to reproduce the evaluation metrics result on BDD like mMOTA etc.

I am facing the same issue now. The testing pipeline seems to work fine but the results do not match the ones in the paper. Did you manage to find a solution? @Raspberry-beans