Hugging Face's logo

mgiardinetti
/
layout-parser-newspapers

Image Segmentation
Transformers
TensorBoard
English
French
humanities
history
newspapers
Inference Endpoints
Model card Files Metrics Community
mgiardinetti commited on
Commit
20169a3
·
verified ·
1 Parent(s): c33420e

Upload config.json

Browse files
Files changed (1) hide show
  1. config.json +404 -0
config.json ADDED
@@ -0,0 +1,404 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "CUDNN_BENCHMARK": false,
3
+ "DATALOADER": {
4
+ "ASPECT_RATIO_GROUPING": true,
5
+ "FILTER_EMPTY_ANNOTATIONS": true,
6
+ "NUM_WORKERS": 4,
7
+ "REPEAT_THRESHOLD": 0,
8
+ "SAMPLER_TRAIN": "TrainingSampler"
9
+ },
10
+ "DATASETS": {
11
+ "PRECOMPUTED_PROPOSAL_TOPK_TEST": 1000,
12
+ "PRECOMPUTED_PROPOSAL_TOPK_TRAIN": 2000,
13
+ "PROPOSAL_FILES_TEST": [],
14
+ "PROPOSAL_FILES_TRAIN": [],
15
+ "TEST": [
16
+ "modele-val"
17
+ ],
18
+ "TRAIN": [
19
+ "modele-train"
20
+ ]
21
+ },
22
+ "GLOBAL": {
23
+ "HACK": 1
24
+ },
25
+ "INPUT": {
26
+ "CROP": {
27
+ "ENABLED": false,
28
+ "SIZE": [
29
+ 0.9,
30
+ 0.9
31
+ ],
32
+ "TYPE": "relative_range"
33
+ },
34
+ "FORMAT": "BGR",
35
+ "MASK_FORMAT": "polygon",
36
+ "MAX_SIZE_TEST": 1333,
37
+ "MAX_SIZE_TRAIN": 1333,
38
+ "MIN_SIZE_TEST": 800,
39
+ "MIN_SIZE_TRAIN": [
40
+ 640,
41
+ 672,
42
+ 704,
43
+ 736,
44
+ 768,
45
+ 800
46
+ ],
47
+ "MIN_SIZE_TRAIN_SAMPLING": "choice",
48
+ "RANDOM_FLIP": "horizontal"
49
+ },
50
+ "MODEL": {
51
+ "ANCHOR_GENERATOR": {
52
+ "ANGLES": [
53
+ [
54
+ -90,
55
+ 0,
56
+ 90
57
+ ]
58
+ ],
59
+ "ASPECT_RATIOS": [
60
+ [
61
+ 0.5,
62
+ 1,
63
+ 2
64
+ ]
65
+ ],
66
+ "NAME": "DefaultAnchorGenerator",
67
+ "OFFSET": 0,
68
+ "SIZES": [
69
+ [
70
+ 32
71
+ ],
72
+ [
73
+ 64
74
+ ],
75
+ [
76
+ 128
77
+ ],
78
+ [
79
+ 256
80
+ ],
81
+ [
82
+ 512
83
+ ]
84
+ ]
85
+ },
86
+ "BACKBONE": {
87
+ "FREEZE_AT": 2,
88
+ "NAME": "build_resnet_fpn_backbone"
89
+ },
90
+ "DEVICE": "cuda",
91
+ "FPN": {
92
+ "FUSE_TYPE": "sum",
93
+ "IN_FEATURES": [
94
+ "res2",
95
+ "res3",
96
+ "res4",
97
+ "res5"
98
+ ],
99
+ "NORM": "",
100
+ "OUT_CHANNELS": 256
101
+ },
102
+ "KEYPOINT_ON": false,
103
+ "LOAD_PROPOSALS": false,
104
+ "MASK_ON": true,
105
+ "META_ARCHITECTURE": "GeneralizedRCNN",
106
+ "PANOPTIC_FPN": {
107
+ "COMBINE": {
108
+ "ENABLED": true,
109
+ "INSTANCES_CONFIDENCE_THRESH": 0.5,
110
+ "OVERLAP_THRESH": 0.5,
111
+ "STUFF_AREA_LIMIT": 4096
112
+ },
113
+ "INSTANCE_LOSS_WEIGHT": 1
114
+ },
115
+ "PIXEL_MEAN": [
116
+ 103.53,
117
+ 116.28,
118
+ 123.675
119
+ ],
120
+ "PIXEL_STD": [
121
+ 1,
122
+ 1,
123
+ 1
124
+ ],
125
+ "PROPOSAL_GENERATOR": {
126
+ "MIN_SIZE": 0,
127
+ "NAME": "RPN"
128
+ },
129
+ "RESNETS": {
130
+ "DEFORM_MODULATED": false,
131
+ "DEFORM_NUM_GROUPS": 1,
132
+ "DEFORM_ON_PER_STAGE": [
133
+ false,
134
+ false,
135
+ false,
136
+ false
137
+ ],
138
+ "DEPTH": 50,
139
+ "NORM": "FrozenBN",
140
+ "NUM_GROUPS": 1,
141
+ "OUT_FEATURES": [
142
+ "res2",
143
+ "res3",
144
+ "res4",
145
+ "res5"
146
+ ],
147
+ "RES2_OUT_CHANNELS": 256,
148
+ "RES5_DILATION": 1,
149
+ "STEM_OUT_CHANNELS": 64,
150
+ "STRIDE_IN_1X1": true,
151
+ "WIDTH_PER_GROUP": 64
152
+ },
153
+ "RETINANET": {
154
+ "BBOX_REG_LOSS_TYPE": "smooth_l1",
155
+ "BBOX_REG_WEIGHTS": [
156
+ 1,
157
+ 1,
158
+ 1,
159
+ 1
160
+ ],
161
+ "FOCAL_LOSS_ALPHA": 0.25,
162
+ "FOCAL_LOSS_GAMMA": 2,
163
+ "IN_FEATURES": [
164
+ "p3",
165
+ "p4",
166
+ "p5",
167
+ "p6",
168
+ "p7"
169
+ ],
170
+ "IOU_LABELS": [
171
+ 0,
172
+ -1,
173
+ 1
174
+ ],
175
+ "IOU_THRESHOLDS": [
176
+ 0.4,
177
+ 0.5
178
+ ],
179
+ "NMS_THRESH_TEST": 0.5,
180
+ "NORM": "",
181
+ "NUM_CLASSES": 80,
182
+ "NUM_CONVS": 4,
183
+ "PRIOR_PROB": 0.01,
184
+ "SCORE_THRESH_TEST": 0.05,
185
+ "SMOOTH_L1_LOSS_BETA": 0.1,
186
+ "TOPK_CANDIDATES_TEST": 1000
187
+ },
188
+ "ROI_BOX_CASCADE_HEAD": {
189
+ "BBOX_REG_WEIGHTS": [
190
+ [
191
+ 10,
192
+ 10,
193
+ 5,
194
+ 5
195
+ ],
196
+ [
197
+ 20,
198
+ 20,
199
+ 10,
200
+ 10
201
+ ],
202
+ [
203
+ 30,
204
+ 30,
205
+ 15,
206
+ 15
207
+ ]
208
+ ],
209
+ "IOUS": [
210
+ 0.5,
211
+ 0.6,
212
+ 0.7
213
+ ]
214
+ },
215
+ "ROI_BOX_HEAD": {
216
+ "BBOX_REG_LOSS_TYPE": "smooth_l1",
217
+ "BBOX_REG_LOSS_WEIGHT": 1,
218
+ "BBOX_REG_WEIGHTS": [
219
+ 10,
220
+ 10,
221
+ 5,
222
+ 5
223
+ ],
224
+ "CLS_AGNOSTIC_BBOX_REG": false,
225
+ "CONV_DIM": 256,
226
+ "FC_DIM": 1024,
227
+ "NAME": "FastRCNNConvFCHead",
228
+ "NORM": "",
229
+ "NUM_CONV": 0,
230
+ "NUM_FC": 2,
231
+ "POOLER_RESOLUTION": 7,
232
+ "POOLER_SAMPLING_RATIO": 0,
233
+ "POOLER_TYPE": "ROIAlignV2",
234
+ "SMOOTH_L1_BETA": 0,
235
+ "TRAIN_ON_PRED_BOXES": false
236
+ },
237
+ "ROI_HEADS": {
238
+ "BATCH_SIZE_PER_IMAGE": 512,
239
+ "IN_FEATURES": [
240
+ "p2",
241
+ "p3",
242
+ "p4",
243
+ "p5"
244
+ ],
245
+ "IOU_LABELS": [
246
+ 0,
247
+ 1
248
+ ],
249
+ "IOU_THRESHOLDS": [
250
+ 0.5
251
+ ],
252
+ "NAME": "StandardROIHeads",
253
+ "NMS_THRESH_TEST": 0.5,
254
+ "NUM_CLASSES": 2,
255
+ "POSITIVE_FRACTION": 0.25,
256
+ "PROPOSAL_APPEND_GT": true,
257
+ "SCORE_THRESH_TEST": 0.05
258
+ },
259
+ "ROI_KEYPOINT_HEAD": {
260
+ "CONV_DIMS": [
261
+ 512,
262
+ 512,
263
+ 512,
264
+ 512,
265
+ 512,
266
+ 512,
267
+ 512,
268
+ 512
269
+ ],
270
+ "LOSS_WEIGHT": 1,
271
+ "MIN_KEYPOINTS_PER_IMAGE": 1,
272
+ "NAME": "KRCNNConvDeconvUpsampleHead",
273
+ "NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS": true,
274
+ "NUM_KEYPOINTS": 17,
275
+ "POOLER_RESOLUTION": 14,
276
+ "POOLER_SAMPLING_RATIO": 0,
277
+ "POOLER_TYPE": "ROIAlignV2"
278
+ },
279
+ "ROI_MASK_HEAD": {
280
+ "CLS_AGNOSTIC_MASK": false,
281
+ "CONV_DIM": 256,
282
+ "NAME": "MaskRCNNConvUpsampleHead",
283
+ "NORM": "",
284
+ "NUM_CONV": 4,
285
+ "POOLER_RESOLUTION": 14,
286
+ "POOLER_SAMPLING_RATIO": 0,
287
+ "POOLER_TYPE": "ROIAlignV2"
288
+ },
289
+ "RPN": {
290
+ "BATCH_SIZE_PER_IMAGE": 256,
291
+ "BBOX_REG_LOSS_TYPE": "smooth_l1",
292
+ "BBOX_REG_LOSS_WEIGHT": 1,
293
+ "BBOX_REG_WEIGHTS": [
294
+ 1,
295
+ 1,
296
+ 1,
297
+ 1
298
+ ],
299
+ "BOUNDARY_THRESH": -1,
300
+ "HEAD_NAME": "StandardRPNHead",
301
+ "IN_FEATURES": [
302
+ "p2",
303
+ "p3",
304
+ "p4",
305
+ "p5",
306
+ "p6"
307
+ ],
308
+ "IOU_LABELS": [
309
+ 0,
310
+ -1,
311
+ 1
312
+ ],
313
+ "IOU_THRESHOLDS": [
314
+ 0.3,
315
+ 0.7
316
+ ],
317
+ "LOSS_WEIGHT": 1,
318
+ "NMS_THRESH": 0.7,
319
+ "POSITIVE_FRACTION": 0.5,
320
+ "POST_NMS_TOPK_TEST": 1000,
321
+ "POST_NMS_TOPK_TRAIN": 1000,
322
+ "PRE_NMS_TOPK_TEST": 1000,
323
+ "PRE_NMS_TOPK_TRAIN": 2000,
324
+ "SMOOTH_L1_BETA": 0
325
+ },
326
+ "SEM_SEG_HEAD": {
327
+ "COMMON_STRIDE": 4,
328
+ "CONVS_DIM": 128,
329
+ "IGNORE_VALUE": 255,
330
+ "IN_FEATURES": [
331
+ "p2",
332
+ "p3",
333
+ "p4",
334
+ "p5"
335
+ ],
336
+ "LOSS_WEIGHT": 1,
337
+ "NAME": "SemSegFPNHead",
338
+ "NORM": "GN",
339
+ "NUM_CLASSES": 54
340
+ },
341
+ "WEIGHTS": "/content/drive/MyDrive/layoutparser/modele/modele3_NP?/model_final.pth"
342
+ },
343
+ "OUTPUT_DIR": "/content/drive/MyDrive/layoutparser/modele",
344
+ "SEED": -1,
345
+ "SOLVER": {
346
+ "AMP": {
347
+ "ENABLED": false
348
+ },
349
+ "BASE_LR": 0.00025,
350
+ "BIAS_LR_FACTOR": 1,
351
+ "CHECKPOINT_PERIOD": 50,
352
+ "CLIP_GRADIENTS": {
353
+ "CLIP_TYPE": "value",
354
+ "CLIP_VALUE": 1,
355
+ "ENABLED": false,
356
+ "NORM_TYPE": 2
357
+ },
358
+ "GAMMA": 0.1,
359
+ "IMS_PER_BATCH": 2,
360
+ "LR_SCHEDULER_NAME": "WarmupMultiStepLR",
361
+ "MAX_ITER": 300,
362
+ "MOMENTUM": 0.9,
363
+ "NESTEROV": false,
364
+ "REFERENCE_WORLD_SIZE": 0,
365
+ "STEPS": [
366
+ 210000,
367
+ 250000
368
+ ],
369
+ "WARMUP_FACTOR": 0.001,
370
+ "WARMUP_ITERS": 1000,
371
+ "WARMUP_METHOD": "linear",
372
+ "WEIGHT_DECAY": 0.0001,
373
+ "WEIGHT_DECAY_BIAS": 0.0001,
374
+ "WEIGHT_DECAY_NORM": 0
375
+ },
376
+ "TEST": {
377
+ "AUG": {
378
+ "ENABLED": false,
379
+ "FLIP": true,
380
+ "MAX_SIZE": 4000,
381
+ "MIN_SIZES": [
382
+ 400,
383
+ 500,
384
+ 600,
385
+ 700,
386
+ 800,
387
+ 900,
388
+ 1000,
389
+ 1100,
390
+ 1200
391
+ ]
392
+ },
393
+ "DETECTIONS_PER_IMAGE": 100,
394
+ "EVAL_PERIOD": 0,
395
+ "EXPECTED_RESULTS": [],
396
+ "KEYPOINT_OKS_SIGMAS": [],
397
+ "PRECISE_BN": {
398
+ "ENABLED": false,
399
+ "NUM_ITER": 200
400
+ }
401
+ },
402
+ "VERSION": 2,
403
+ "VIS_PERIOD": 0
404
+ }