darknet-metadata.json 20 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480
  1. [
  2. {
  3. "name": "avgpool",
  4. "category": "Pool"
  5. },
  6. {
  7. "name": "batchnorm",
  8. "category": "Normalization"
  9. },
  10. {
  11. "name": "connected",
  12. "category": "Layer",
  13. "attributes": [
  14. { "name": "output", "type": "int32", "visible": false, "default": 1 },
  15. { "name": "activation", "type": "string", "default": "logistic", "description": "options are: LOGISTIC, RELU, RELIE, LINEAR, RAMP, TANH, PLSE, LEAKY, ELU, LOGGY, STAIR, HARDTAN, LHTAN, SELU, SWISH, MISH" },
  16. { "name": "batch_normalize", "type": "int32", "default": 0 }
  17. ]
  18. },
  19. {
  20. "name": "conv_lstm",
  21. "category": "Layer",
  22. "attributes": [
  23. { "name": "batch_normalize", "type": "int32", "default": 0 },
  24. { "name": "activation", "type": "string", "default": "linear", "description": "options are: LOGISTIC, RELU, RELIE, LINEAR, RAMP, TANH, PLSE, LEAKY, ELU, LOGGY, STAIR, HARDTAN, LHTAN, SELU, SWISH, MISH" },
  25. { "name": "size", "type": "int32", "default": 3 },
  26. { "name": "stride", "type": "int32", "default": 1 },
  27. { "name": "dilation", "default": 1 },
  28. { "name": "groups", "type": "int32", "default": 1 },
  29. { "name": "padding", "type": "int32", "default": 0 },
  30. { "name": "pad", "type": "int32", "default": 0 },
  31. { "name": "xnor", "type": "int32", "default": 0 },
  32. { "name": "shortcut", "default": 0 },
  33. { "name": "output", "type": "int32", "default": 1 },
  34. { "name": "state_constrain", "type": "int32", "default": 16 },
  35. { "name": "peephole", "type": "int32", "default": 0 }
  36. ]
  37. },
  38. {
  39. "name": "convolutional",
  40. "category": "Layer",
  41. "attributes": [
  42. { "name": "filters", "type": "int32", "default": 1 },
  43. { "name": "size", "type": "int32", "default": 1 },
  44. { "name": "stride", "type": "int32", "default": 1 },
  45. { "name": "stride_x", "type": "int32", "default": -1 },
  46. { "name": "stride_y", "type": "int32", "default": -1 },
  47. { "name": "groups", "type": "int32", "default": 1 },
  48. { "name": "padding", "type": "int32", "default": 0 },
  49. { "name": "pad", "type": "int32", "default": 0 },
  50. { "name": "dilation", "default": 1 },
  51. { "name": "share_index", "default": -1000000000 },
  52. { "name": "binary", "type": "int32", "default": 0 },
  53. { "name": "xnor", "type": "int32", "default": 0 },
  54. { "name": "bin_output", "type": "int32", "default": 0 },
  55. { "name": "flipped", "type": "int32", "default": 0 },
  56. { "name": "dot", "type": "float32", "default": 0 },
  57. { "name": "batch_normalize", "type": "int32", "default": 0 },
  58. { "name": "activation", "type": "string", "default": "logistic", "description": "options are: LOGISTIC, RELU, RELIE, LINEAR, RAMP, TANH, PLSE, LEAKY, ELU, LOGGY, STAIR, HARDTAN, LHTAN, SELU, SWISH, MISH" }
  59. ]
  60. },
  61. {
  62. "name": "cost",
  63. "attributes": [
  64. { "name": "type", "type": "string", "default": "sse" },
  65. { "name": "scale", "type": "float32", "default": 1 },
  66. { "name": "ratio", "type": "float32", "default": 0 }
  67. ]
  68. },
  69. {
  70. "name": "crnn",
  71. "category": "Layer",
  72. "attributes": [
  73. { "name": "batch_normalize", "type": "int32", "default": 0 },
  74. { "name": "activation", "type": "string", "default": "logistic" },
  75. { "name": "dilation", "default": 1 },
  76. { "name": "padding", "default": 0 },
  77. { "name": "pad", "type": "int32", "default": 0 },
  78. { "name": "groups", "type": "int32", "default": 1 },
  79. { "name": "xnor", "type": "int32", "default": 0 },
  80. { "name": "shortcut", "type": "int32", "default": 0 },
  81. { "name": "output_filters", "default": 1 },
  82. { "name": "hidden_filters", "default": 1 }
  83. ]
  84. },
  85. {
  86. "name": "crop",
  87. "category": "Shape",
  88. "attributes": [
  89. { "name": "crop_height", "type": "int32", "default": 1 },
  90. { "name": "crop_width", "type": "int32", "default": 1 },
  91. { "name": "flip", "type": "int32", "default": 0 },
  92. { "name": "exposure", "type": "float32", "default": 1 },
  93. { "name": "saturation", "type": "float32", "default": 1 },
  94. { "name": "angle", "type": "float32", "default": 0 },
  95. { "name": "noadjust", "default": 0 },
  96. { "name": "shift", "type": "float32", "default": 0 }
  97. ]
  98. },
  99. {
  100. "name": "deconvolutional",
  101. "category": "Layer",
  102. "attributes": [
  103. { "name": "filters", "type": "int32", "visible": false, "default": 1 },
  104. { "name": "size", "type": "int32", "default": 1 },
  105. { "name": "stride", "type": "int32", "default": 1 },
  106. { "name": "padding", "type": "int32", "default": 0 },
  107. { "name": "pad", "type": "int32", "default": 0 },
  108. { "name": "batch_normalize", "type": "int32", "default": 0 },
  109. { "name": "activation", "type": "string", "default": "logistic", "description": "options are: LOGISTIC, RELU, RELIE, LINEAR, RAMP, TANH, PLSE, LEAKY, ELU, LOGGY, STAIR, HARDTAN, LHTAN, SELU, SWISH, MISH" }
  110. ]
  111. },
  112. {
  113. "name": "detection",
  114. "category": "Layer",
  115. "attributes": [
  116. { "name": "classes", "type": "int32", "default": 1 },
  117. { "name": "coord", "type": "int32", "default": 1 },
  118. { "name": "num", "type": "int32", "default": 1 },
  119. { "name": "jitter", "type": "float32", "default": 0.2 },
  120. { "name": "coord_scale", "type": "float32", "default": 1 },
  121. { "name": "object_scale", "type": "float32", "default": 1 },
  122. { "name": "noobject_scale", "type": "float32", "default": 1 },
  123. { "name": "class_scale", "type": "float32", "default": 1 },
  124. { "name": "forced", "type": "int32", "default": 0 },
  125. { "name": "side", "type": "int32", "default": 7 },
  126. { "name": "softmax", "type": "int32", "default": 0 },
  127. { "name": "sqrt", "type": "int32", "default": 0 },
  128. { "name": "max", "type": "int32", "default": 30 },
  129. { "name": "rescore", "type": "int32", "default": 0 },
  130. { "name": "random", "type": "int32", "default": 0 },
  131. { "name": "reorg", "type": "int32", "default": 0 }
  132. ]
  133. },
  134. {
  135. "name": "dropout",
  136. "category": "Dropout",
  137. "attributes": [
  138. { "name": "probability", "type": "float32", "default": 0.5 },
  139. { "name": "dropblock", "type": "int32", "default": 0 },
  140. { "name": "dropblock_size_rel", "type": "float32", "default": 0 },
  141. { "name": "dropblock_size_abs ", "type": "int32", "default": 7 }
  142. ]
  143. },
  144. {
  145. "name": "elu",
  146. "category": "Activation"
  147. },
  148. {
  149. "name": "gaussian_yolo",
  150. "category": "Layer",
  151. "attributes": [
  152. { "name": "classes", "type": "int32", "default": 20 },
  153. { "name": "num", "type": "int32", "default": 1 },
  154. { "name": "mask", "type": "string", "default": 0 },
  155. { "name": "jitter", "type": "float32", "default": 0.2 },
  156. { "name": "label_smooth_eps", "type": "float32", "default": 0 },
  157. { "name": "scale_x_y", "type": "float32", "default": 1 },
  158. { "name": "uc_normalizer", "type": "float32", "default": 1 },
  159. { "name": "iou_normalizer", "type": "float32", "default": 0.75 },
  160. { "name": "cls_normalizer", "type": "float32", "default": 1 },
  161. { "name": "iou_loss", "type": "string", "default": "mse", "description": "options are: mse, giou, diou, and ciou" },
  162. { "name": "max", "default": 90 },
  163. { "name": "ignore_thresh", "type": "float32", "default": 0.5 },
  164. { "name": "truth_thresh", "default": 1 },
  165. { "name": "iou_thresh", "type": "float32", "default": 1, "description": "recommended to use iou_thresh=0.213" },
  166. { "name": "random", "type": "int32", "default": 0 },
  167. { "name": "map", "type": "string", "default": 0 },
  168. { "name": "beta_nms", "type": "float32", "default": 0.6 },
  169. { "name": "nms_kind", "type": "string", "default": "default", "description": "options are: greedynms, diounms, cornersnms, or defaultnms" },
  170. { "name": "anchors", "type": "string", "default": 0 },
  171. { "name": "yolo_point", "type": "string", "default": "center", "description": "options are: center, left_top, and right_bottom" }
  172. ]
  173. },
  174. {
  175. "name": "gru",
  176. "category": "Layer",
  177. "attributes": [
  178. { "name": "output", "type": "int32", "visible": false, "default": 1 },
  179. { "name": "batch_normalize", "type": "int32", "default": 0 }
  180. ]
  181. },
  182. {
  183. "name": "h_swish",
  184. "category": "Activation"
  185. },
  186. {
  187. "name": "hardtan",
  188. "category": "Activation"
  189. },
  190. {
  191. "name": "leaky",
  192. "category": "Activation"
  193. },
  194. {
  195. "name": "lhtan",
  196. "category": "Activation"
  197. },
  198. {
  199. "name": "linear",
  200. "category": "Activation"
  201. },
  202. {
  203. "name": "local",
  204. "category": "Layer",
  205. "attributes": [
  206. { "name": "filters", "type": "int32", "visible": false, "default": 1 },
  207. { "name": "size", "type": "int32", "default": 1 },
  208. { "name": "stride", "type": "int32", "default": 1 },
  209. { "name": "padding", "type": "int32", "default": 0 },
  210. { "name": "pad", "type": "int32", "default": 0 },
  211. { "name": "activation", "type": "string", "default": "logistic", "description": "options are: LOGISTIC, RELU, RELIE, LINEAR, RAMP, TANH, PLSE, LEAKY, ELU, LOGGY, STAIR, HARDTAN, LHTAN, SELU, SWISH, MISH" }
  212. ]
  213. },
  214. {
  215. "name": "loggy",
  216. "category": "Activation"
  217. },
  218. {
  219. "name": "logistic",
  220. "category": "Activation"
  221. },
  222. {
  223. "name": "lstm",
  224. "category": "Layer",
  225. "attributes": [
  226. { "name": "output", "type": "int32", "visible": false, "default": 1 },
  227. { "name": "batch_normalize", "type": "int32", "default": 0 }
  228. ]
  229. },
  230. {
  231. "name": "maxpool",
  232. "category": "Pool",
  233. "attributes": [
  234. { "name": "stride", "type": "int32", "default": 1 },
  235. { "name": "stride_x", "type": "int32", "default": 1 },
  236. { "name": "stride_y", "type": "int32", "default": 1 },
  237. { "name": "size", "type": "int32", "default": 1 },
  238. { "name": "padding", "type": "int32", "default": 0 },
  239. { "name": "maxpool_depth", "type": "int32", "default": 0 },
  240. { "name": "out_channels", "default": 1 },
  241. { "name": "antialiasing", "type": "int32", "default": 0 }
  242. ]
  243. },
  244. {
  245. "name": "mish",
  246. "category": "Activation"
  247. },
  248. {
  249. "name": "net",
  250. "attributes": [
  251. { "name": "batch", "type": "int32", "default": 1 },
  252. { "name": "max_batches", "type": "int32", "default": 0, "description": "Limits the maximum number of iterations" },
  253. { "name": "learning_rate", "type": "float32", "default": 0.001 },
  254. { "name": "momentum", "type": "float32", "default": 0.9 },
  255. { "name": "decay", "type": "float32", "default": 0.0001 },
  256. { "name": "subdivisions", "type": "int32", "default": 1, "description": "In concert with batch property, this greatly affect memory usage, minimal working number is recommended" },
  257. { "name": "time_steps", "type": "int32", "default": 1 },
  258. { "name": "notruth", "type": "int32", "default": 0 },
  259. { "name": "random", "type": "int32", "default": 0 },
  260. { "name": "adam", "type": "int32", "default": 0 },
  261. { "name": "B1", "type": "float32", "default": 0.9 },
  262. { "name": "B2", "type": "float32", "default": 0.999 },
  263. { "name": "eps", "type": "float32", "default": 1e-7 },
  264. { "name": "height", "type": "int32", "default": 0 },
  265. { "name": "width", "type": "int32", "default": 0 },
  266. { "name": "channels", "type": "int32", "default": 0 },
  267. { "name": "inputs", "type": "int32" },
  268. { "name": "max_crop", "type": "int32" },
  269. { "name": "min_crop", "type": "int32" },
  270. { "name": "max_ratio", "type": "float32" },
  271. { "name": "min_ratio", "type": "float32" },
  272. { "name": "center", "type": "int32", "default": 0 },
  273. { "name": "clip", "type": "int32", "default": 0 },
  274. { "name": "angle", "type": "float32", "default": 0 },
  275. { "name": "aspect", "type": "float32", "default": 1 },
  276. { "name": "saturation", "type": "float32", "default": 1 },
  277. { "name": "exposure", "type": "float32", "default": 1 },
  278. { "name": "hue", "type": "float32", "default": 0 },
  279. { "name": "power", "type": "float32", "default": 4 },
  280. { "name": "flip", "type": "int32", "default": 1, "description": "Enables augmentation method: horizontal flip" },
  281. { "name": "blur", "type": "int32", "default": 0, "description": "Enables augmentation method: backgound blurring" },
  282. { "name": "mixup", "type": "int32", "default": 0, "description": "Enables augmentation method: images mixup" },
  283. { "name": "cutmix", "type": "int32", "default": 0, "description": "Enables augmentation method: images cutmix" },
  284. { "name": "mosaic", "type": "int32", "default": 0, "description": "Enables augmentation method: images mosaicing" },
  285. { "name": "letter_box", "type": "int32", "default": 0, "description": "Enables letter-box resizing (keeping the aspect ratio)" },
  286. { "name": "policy", "type": "string", "default": "constant" },
  287. { "name": "burn_in", "type": "int32", "default": 0, "description": "Is used for MAP calculation: permit a minimal number of iteration before first MAP check" },
  288. { "name": "letter_box", "type": "int32", "default": 0 },
  289. { "name": "optimized_memory", "type": "int32", "default": 0, "description": "can offload memory from GPU into CPU at the cost of speed, 3 options are possible please look at: https://github.com/AlexeyAB/darknet/issues/4386" },
  290. { "name": "workspace_size_limit_MB", "type": "float32", "default": 1024 }
  291. ]
  292. },
  293. {
  294. "name": "norm_chan",
  295. "category": "Activation"
  296. },
  297. {
  298. "name": "norm_chan_softmax",
  299. "category": "Activation"
  300. },
  301. {
  302. "name": "normalization",
  303. "category": "Normalization",
  304. "attributes": [
  305. { "name": "alpha", "type": "float32", "default": 0.0001 },
  306. { "name": "beta", "type": "float32", "default": 0.75 },
  307. { "name": "kappa", "type": "float32", "default": 1 },
  308. { "name": "size", "default": 5 }
  309. ]
  310. },
  311. {
  312. "name": "plse",
  313. "category": "Activation"
  314. },
  315. {
  316. "name": "ramp",
  317. "category": "Activation"
  318. },
  319. {
  320. "name": "region",
  321. "category": "Layer",
  322. "attributes": [
  323. { "name": "classes", "type": "int32", "default": 20 },
  324. { "name": "coord", "default": 4 },
  325. { "name": "num", "default": 1 },
  326. { "name": "mask", "type": "string", "default": 0 },
  327. { "name": "jitter", "type": "float32", "default": 0.2 },
  328. { "name": "classfix", "type": "int32", "default": 0 },
  329. { "name": "coord_scale", "type": "float32", "default": 1 },
  330. { "name": "object_scale", "type": "float32", "default": 1 },
  331. { "name": "noobject_scale", "type": "float32", "default": 1 },
  332. { "name": "mask_scale", "type": "float32", "default": 1 },
  333. { "name": "class_scale", "type": "float32", "default": 1 },
  334. { "name": "bias_match", "type": "int32", "default": 0 },
  335. { "name": "focal_loss", "type": "int32", "default": 0 },
  336. { "name": "max", "type": "int32", "default": 90 },
  337. { "name": "softmax", "type": "int32", "default": 0 },
  338. { "name": "rescore", "type": "int32", "default": 0 },
  339. { "name": "thresh", "type": "float32", "default": 0.5 },
  340. { "name": "random", "type": "int32", "default": 0 },
  341. { "name": "map", "type": "string", "default": 0 },
  342. { "name": "tree", "type": "string", "default": 0 },
  343. { "name": "anchors", "type": "string", "default": 0 },
  344. { "name": "absolute", "default": 0 },
  345. { "name": "log", "default": 0 },
  346. { "name": "sqrt", "default": 0 }
  347. ]
  348. },
  349. {
  350. "name": "relie",
  351. "category": "Activation"
  352. },
  353. {
  354. "name": "relu",
  355. "category": "Activation"
  356. },
  357. {
  358. "name": "relu6",
  359. "category": "Activation"
  360. },
  361. {
  362. "name": "reorg",
  363. "category": "Shape",
  364. "attributes": [
  365. { "name": "stride", "default": 1 },
  366. { "name": "reverse", "type": "int32", "default": 0 }
  367. ]
  368. },
  369. {
  370. "name": "rnn",
  371. "category": "Layer",
  372. "attributes": [
  373. { "name": "output", "type": "int32", "visible": false },
  374. { "name": "hidden", "visible": false, "default": 1 },
  375. { "name": "activation", "type": "string", "default": "logistic", "description": "options are: LOGISTIC, RELU, RELIE, LINEAR, RAMP, TANH, PLSE, LEAKY, ELU, LOGGY, STAIR, HARDTAN, LHTAN, SELU, SWISH, MISH" },
  376. { "name": "groups", "type": "int32", "default": 1 },
  377. { "name": "xnor", "type": "int32", "default": 0 },
  378. { "name": "shortcut", "default": 0 },
  379. { "name": "logistic", "default": 0 },
  380. { "name": "batch_normalize", "type": "int32", "default": 0 }
  381. ]
  382. },
  383. {
  384. "name": "route",
  385. "category": "Tensor",
  386. "attributes": [
  387. { "name": "groups_id", "type": "int32", "default": 0 },
  388. { "name": "groups", "type": "int32", "default": 1 }
  389. ]
  390. },
  391. {
  392. "name": "sam",
  393. "category": "Tensor",
  394. "attributes": [
  395. { "name": "activation", "type": "string", "default": "linear", "description": "options are: LOGISTIC, RELU, RELIE, LINEAR, RAMP, TANH, PLSE, LEAKY, ELU, LOGGY, STAIR, HARDTAN, LHTAN, SELU" },
  396. { "name": "from", "description": "This params link the layer to another one, the index of the layer is either positive in which case it's a direct address, if negative it's relative to the layer position" }
  397. ]
  398. },
  399. {
  400. "name": "scale_channels",
  401. "category": "Tensor",
  402. "attributes": [
  403. { "name": "activation", "type": "string", "default": "linear", "description": "options are: LOGISTIC, RELU, RELIE, LINEAR, RAMP, TANH, PLSE, LEAKY, ELU, LOGGY, STAIR, HARDTAN, LHTAN, SELU" },
  404. { "name": "scale_wh", "default": 0 },
  405. { "name": "from", "description": "This params link the layer to another one, the index of the layer is either positive in which case it's a direct address, if negative it's relative to the layer position" }
  406. ]
  407. },
  408. {
  409. "name": "selu",
  410. "category": "Activation"
  411. },
  412. {
  413. "name": "silu",
  414. "category": "Activation"
  415. },
  416. {
  417. "name": "shortcut",
  418. "category": "Tensor",
  419. "attributes": [
  420. { "name": "activation", "type": "string", "default": "logistic", "description": "options are: LOGISTIC, RELU, RELIE, LINEAR, RAMP, TANH, PLSE, LEAKY, ELU, LOGGY, STAIR, HARDTAN, LHTAN, SELU, SWISH, MISH" },
  421. { "name": "assisted_excitation", "default": 0 },
  422. { "name": "from", "description": "This params link the layer to another one, the index of the layer is either positive in which case it's a direct address, if negative it's relative to the layer position" }
  423. ]
  424. },
  425. {
  426. "name": "softmax",
  427. "category": "Activation",
  428. "attributes": [
  429. { "name": "groups", "type": "int32", "default": 1 },
  430. { "name": "temperature", "type": "float32", "default": 1 },
  431. { "name": "tree", "type": "string", "default": 0 },
  432. { "name": "spatial", "type": "int32", "default": 0 },
  433. { "name": "noloss", "type": "int32", "default": 0 }
  434. ]
  435. },
  436. {
  437. "name": "stair",
  438. "category": "Activation"
  439. },
  440. {
  441. "name": "swish",
  442. "category": "Activation"
  443. },
  444. {
  445. "name": "tanh",
  446. "category": "Activation"
  447. },
  448. {
  449. "name": "upsample",
  450. "category": "Data",
  451. "attributes": [
  452. { "name": "stride", "type": "int32", "default": 2 },
  453. { "name": "scale", "type": "float32", "default": 1 }
  454. ]
  455. },
  456. {
  457. "name": "yolo",
  458. "category": "Layer",
  459. "attributes": [
  460. { "name": "classes", "type": "int32", "default": 20 },
  461. { "name": "num", "type": "int32", "default": 1 },
  462. { "name": "mask", "type": "int32[]", "default": 0 },
  463. { "name": "jitter", "type": "float32", "default": 0.2 },
  464. { "name": "label_smooth_eps", "type": "float32", "default": 0 },
  465. { "name": "scale_x_y", "type": "float32", "default": 1 },
  466. { "name": "iou_normalizer", "type": "float32", "default": 0.75 },
  467. { "name": "cls_normalizer", "type": "float32", "default": 1 },
  468. { "name": "iou_loss", "type": "string", "default": "mse", "description": "options are: mse, giou, diou, and ciou" },
  469. { "name": "focal_loss", "type": "int32", "default": 0 },
  470. { "name": "max", "type": "int32", "default": 90 },
  471. { "name": "ignore_thresh", "type": "float32", "default": 0.5 },
  472. { "name": "truth_thresh", "type": "float32", "default": 1 },
  473. { "name": "iou_thresh", "type": "float32", "default": 1, "description": "recommended to use iou_thresh=0.213" },
  474. { "name": "random", "type": "int32", "default": 0 },
  475. { "name": "map", "type": "string", "default": 0 },
  476. { "name": "nms_kind", "type": "string", "default": "default", "description": "options are: greedynms, diounms, cornersnms, or defaultnms" },
  477. { "name": "anchors", "type": "int32[]", "default": 0 }
  478. ]
  479. }
  480. ]