VBoussot commited on
Commit
47ffad6
·
verified ·
1 Parent(s): 8acfc94

Give each layer its receptive field in models.json, and keep one DINOv2 model

Browse files

models.json gives every model a per-layer receptive field (fov: [...], null for a model that sees whole images only) and the multiple its input must divide by, in place of the FOV formula. Dino/DinoV2_Small.pt returns the patch tokens of the whole slice (blocks 3, 6, 9, 12), the feature DINO-Reg registers with; the class-token export and DinoV2_Small_PatchTokens.pt go.

Dino/DinoV2_Small.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:e8f4eb22390d10f80df72e801c483db6ad528922c4b2428149f5d81f28edfc43
3
- size 88590113
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:02a98afcb66bba1f38981ae66e8ed0931de75b2a80baa240d6bc3a8c91732d40
3
+ size 88586017
Dino/DinoV2_Small_PatchTokens.pt DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:5372d2cfd4089f00d81c1b8fca8ba39a5e7ebc5484e6f78cdda8465b634517ee
3
- size 88601647
 
 
 
 
models.json CHANGED
@@ -1,250 +1,337 @@
1
  {
2
  "Anatomix/Anatomix.pt": {
3
- "fov": {
4
- "formula": "Global(Static mode)"
5
- },
6
  "dimension": "3",
7
- "numberofchannels": "1"
 
 
8
  },
9
  "Anatomix/AnatomixDevViT.pt": {
10
- "fov": {
11
- "formula": "128"
12
- },
13
  "dimension": "3",
14
- "numberofchannels": "1"
 
 
 
15
  },
16
  "Dino/DinoV2_Small.pt": {
17
- "fov": {
18
- "formula": "14"
19
- },
20
  "dimension": "2",
21
- "numberofchannels": "3"
 
22
  },
23
  "MIND/R1D1_2D.pt": {
24
- "fov": {
25
- "formula": "2*r*d+1",
26
- "r": 1,
27
- "d": 1,
28
- "value": 3
29
- },
30
  "dimension": "2",
31
- "numberofchannels": "1"
 
 
 
32
  },
33
  "MIND/R1D1_3D.pt": {
34
- "fov": {
35
- "formula": "2*r*d+1",
36
- "r": 1,
37
- "d": 1,
38
- "value": 3
39
- },
40
  "dimension": "3",
41
- "numberofchannels": "1"
 
 
 
42
  },
43
  "MIND/R1D2_2D.pt": {
44
- "fov": {
45
- "formula": "2*r*d+1",
46
- "r": 1,
47
- "d": 2,
48
- "value": 5
49
- },
50
  "dimension": "2",
51
- "numberofchannels": "1"
 
 
 
52
  },
53
  "MIND/R1D2_3D.pt": {
54
- "fov": {
55
- "formula": "2*r*d+1",
56
- "r": 1,
57
- "d": 2,
58
- "value": 5
59
- },
60
  "dimension": "3",
61
- "numberofchannels": "1"
 
 
 
62
  },
63
  "MIND/R2D1_2D.pt": {
64
- "fov": {
65
- "formula": "2*r*d+1",
66
- "r": 2,
67
- "d": 1,
68
- "value": 5
69
- },
70
  "dimension": "2",
71
- "numberofchannels": "1"
 
 
 
72
  },
73
  "MIND/R2D1_3D.pt": {
74
- "fov": {
75
- "formula": "2*r*d+1",
76
- "r": 2,
77
- "d": 1,
78
- "value": 5
79
- },
80
  "dimension": "3",
81
- "numberofchannels": "1"
 
 
 
82
  },
83
  "MIND/R2D2_2D.pt": {
84
- "fov": {
85
- "formula": "2*r*d+1",
86
- "r": 2,
87
- "d": 2,
88
- "value": 9
89
- },
90
  "dimension": "2",
91
- "numberofchannels": "1"
 
 
 
92
  },
93
  "MIND/R2D2_3D.pt": {
94
- "fov": {
95
- "formula": "2*r*d+1",
96
- "r": 2,
97
- "d": 2,
98
- "value": 9
99
- },
100
  "dimension": "3",
101
- "numberofchannels": "1"
 
 
 
102
  },
103
  "MRSeg/MRSeg.pt": {
104
- "fov": {
105
- "formula": "2^l+3"
106
- },
107
  "dimension": "3",
108
- "numberofchannels": "1"
 
 
 
 
 
 
 
 
 
 
109
  },
110
  "SAM2.1/SAM2.1_Small.pt": {
111
- "fov": {
112
- "formula": "29"
113
- },
114
  "dimension": "2",
115
- "numberofchannels": "3"
 
 
 
 
 
 
116
  },
117
  "SAM2.1/SAM2.1_Tiny.pt": {
118
- "fov": {
119
- "formula": "29"
120
- },
121
  "dimension": "2",
122
- "numberofchannels": "3"
 
 
 
 
 
 
123
  },
124
  "TS/M258.pt": {
125
- "fov": {
126
- "formula": "2^l+3"
127
- },
128
  "dimension": "3",
129
- "numberofchannels": "1"
 
 
 
 
 
 
 
 
 
 
130
  },
131
  "TS/M291.pt": {
132
- "fov": {
133
- "formula": "2^l+3"
134
- },
135
  "dimension": "3",
136
- "numberofchannels": "1"
 
 
 
 
 
 
 
 
 
 
137
  },
138
  "TS/M292.pt": {
139
- "fov": {
140
- "formula": "2^l+3"
141
- },
142
  "dimension": "3",
143
- "numberofchannels": "1"
 
 
 
 
 
 
 
 
 
 
144
  },
145
  "TS/M293.pt": {
146
- "fov": {
147
- "formula": "2^l+3"
148
- },
149
  "dimension": "3",
150
- "numberofchannels": "1"
 
 
 
 
 
 
 
 
 
 
151
  },
152
  "TS/M294.pt": {
153
- "fov": {
154
- "formula": "2^l+3"
155
- },
156
  "dimension": "3",
157
- "numberofchannels": "1"
 
 
 
 
 
 
 
 
 
 
158
  },
159
  "TS/M295.pt": {
160
- "fov": {
161
- "formula": "2^l+3"
162
- },
163
  "dimension": "3",
164
- "numberofchannels": "1"
 
 
 
 
 
 
 
 
 
 
165
  },
166
  "TS/M297.pt": {
167
- "fov": {
168
- "formula": "2^l+3"
169
- },
170
  "dimension": "3",
171
- "numberofchannels": "1"
 
 
 
 
 
 
 
 
 
172
  },
173
  "TS/M298.pt": {
174
- "fov": {
175
- "formula": "2^l+3"
176
- },
177
  "dimension": "3",
178
- "numberofchannels": "1"
 
 
 
 
 
 
 
 
 
179
  },
180
  "TS/M730.pt": {
181
- "fov": {
182
- "formula": "2^l+3"
183
- },
184
  "dimension": "3",
185
- "numberofchannels": "1"
 
 
 
 
 
 
 
 
 
 
186
  },
187
  "TS/M731.pt": {
188
- "fov": {
189
- "formula": "2^l+3"
190
- },
191
  "dimension": "3",
192
- "numberofchannels": "1"
 
 
 
 
 
 
 
 
 
 
193
  },
194
  "TS/M732.pt": {
195
- "fov": {
196
- "formula": "2^l+3"
197
- },
198
  "dimension": "3",
199
- "numberofchannels": "1"
 
 
 
 
 
 
 
 
 
200
  },
201
  "TS/M733.pt": {
202
- "fov": {
203
- "formula": "2^l+3"
204
- },
205
  "dimension": "3",
206
- "numberofchannels": "1"
 
 
 
 
 
 
 
 
207
  },
208
  "TS/M850.pt": {
209
- "fov": {
210
- "formula": "2^l+3"
211
- },
212
  "dimension": "3",
213
- "numberofchannels": "1"
 
 
 
 
 
 
 
 
 
 
214
  },
215
  "TS/M851.pt": {
216
- "fov": {
217
- "formula": "2^l+3"
218
- },
219
  "dimension": "3",
220
- "numberofchannels": "1"
 
 
 
 
 
 
 
 
 
 
221
  },
222
  "TS/M852.pt": {
223
- "fov": {
224
- "formula": "2^l+3"
225
- },
226
  "dimension": "3",
227
- "numberofchannels": "1"
 
 
 
 
 
 
 
 
 
228
  },
229
  "TS/M853.pt": {
230
- "fov": {
231
- "formula": "2^l+3"
232
- },
233
  "dimension": "3",
234
- "numberofchannels": "1"
 
 
 
 
 
 
 
 
235
  },
236
  "VGG/VGG16.pt": {
237
- "fov": {
238
- "formula": "2^l+3"
239
- },
240
  "dimension": "2",
241
- "numberofchannels": "3"
242
- },
243
- "Dino/DinoV2_Small_PatchTokens.pt": {
244
- "fov": {
245
- "formula": "14"
246
- },
247
- "dimension": "2",
248
- "numberofchannels": "3"
249
  }
250
  }
 
1
  {
2
  "Anatomix/Anatomix.pt": {
 
 
 
3
  "dimension": "3",
4
+ "numberofchannels": "1",
5
+ "fov": null,
6
+ "multiple": 16
7
  },
8
  "Anatomix/AnatomixDevViT.pt": {
 
 
 
9
  "dimension": "3",
10
+ "numberofchannels": "1",
11
+ "fov": [
12
+ 128
13
+ ]
14
  },
15
  "Dino/DinoV2_Small.pt": {
 
 
 
16
  "dimension": "2",
17
+ "numberofchannels": "3",
18
+ "fov": null
19
  },
20
  "MIND/R1D1_2D.pt": {
 
 
 
 
 
 
21
  "dimension": "2",
22
+ "numberofchannels": "1",
23
+ "fov": [
24
+ 3
25
+ ]
26
  },
27
  "MIND/R1D1_3D.pt": {
 
 
 
 
 
 
28
  "dimension": "3",
29
+ "numberofchannels": "1",
30
+ "fov": [
31
+ 3
32
+ ]
33
  },
34
  "MIND/R1D2_2D.pt": {
 
 
 
 
 
 
35
  "dimension": "2",
36
+ "numberofchannels": "1",
37
+ "fov": [
38
+ 5
39
+ ]
40
  },
41
  "MIND/R1D2_3D.pt": {
 
 
 
 
 
 
42
  "dimension": "3",
43
+ "numberofchannels": "1",
44
+ "fov": [
45
+ 5
46
+ ]
47
  },
48
  "MIND/R2D1_2D.pt": {
 
 
 
 
 
 
49
  "dimension": "2",
50
+ "numberofchannels": "1",
51
+ "fov": [
52
+ 5
53
+ ]
54
  },
55
  "MIND/R2D1_3D.pt": {
 
 
 
 
 
 
56
  "dimension": "3",
57
+ "numberofchannels": "1",
58
+ "fov": [
59
+ 5
60
+ ]
61
  },
62
  "MIND/R2D2_2D.pt": {
 
 
 
 
 
 
63
  "dimension": "2",
64
+ "numberofchannels": "1",
65
+ "fov": [
66
+ 9
67
+ ]
68
  },
69
  "MIND/R2D2_3D.pt": {
 
 
 
 
 
 
70
  "dimension": "3",
71
+ "numberofchannels": "1",
72
+ "fov": [
73
+ 9
74
+ ]
75
  },
76
  "MRSeg/MRSeg.pt": {
 
 
 
77
  "dimension": "3",
78
+ "numberofchannels": "1",
79
+ "fov": [
80
+ 5,
81
+ 11,
82
+ 23,
83
+ 47,
84
+ 95,
85
+ 191,
86
+ 191,
87
+ 191
88
+ ]
89
  },
90
  "SAM2.1/SAM2.1_Small.pt": {
 
 
 
91
  "dimension": "2",
92
+ "numberofchannels": "3",
93
+ "fov": [
94
+ 29,
95
+ 29,
96
+ 29,
97
+ 29
98
+ ]
99
  },
100
  "SAM2.1/SAM2.1_Tiny.pt": {
 
 
 
101
  "dimension": "2",
102
+ "numberofchannels": "3",
103
+ "fov": [
104
+ 29,
105
+ 29,
106
+ 29,
107
+ 29
108
+ ]
109
  },
110
  "TS/M258.pt": {
 
 
 
111
  "dimension": "3",
112
+ "numberofchannels": "1",
113
+ "fov": [
114
+ 5,
115
+ 11,
116
+ 23,
117
+ 47,
118
+ 95,
119
+ 191,
120
+ 191,
121
+ 191
122
+ ]
123
  },
124
  "TS/M291.pt": {
 
 
 
125
  "dimension": "3",
126
+ "numberofchannels": "1",
127
+ "fov": [
128
+ 5,
129
+ 11,
130
+ 23,
131
+ 47,
132
+ 95,
133
+ 191,
134
+ 191,
135
+ 191
136
+ ]
137
  },
138
  "TS/M292.pt": {
 
 
 
139
  "dimension": "3",
140
+ "numberofchannels": "1",
141
+ "fov": [
142
+ 5,
143
+ 11,
144
+ 23,
145
+ 47,
146
+ 95,
147
+ 191,
148
+ 191,
149
+ 191
150
+ ]
151
  },
152
  "TS/M293.pt": {
 
 
 
153
  "dimension": "3",
154
+ "numberofchannels": "1",
155
+ "fov": [
156
+ 5,
157
+ 11,
158
+ 23,
159
+ 47,
160
+ 95,
161
+ 191,
162
+ 191,
163
+ 191
164
+ ]
165
  },
166
  "TS/M294.pt": {
 
 
 
167
  "dimension": "3",
168
+ "numberofchannels": "1",
169
+ "fov": [
170
+ 5,
171
+ 11,
172
+ 23,
173
+ 47,
174
+ 95,
175
+ 191,
176
+ 191,
177
+ 191
178
+ ]
179
  },
180
  "TS/M295.pt": {
 
 
 
181
  "dimension": "3",
182
+ "numberofchannels": "1",
183
+ "fov": [
184
+ 5,
185
+ 11,
186
+ 23,
187
+ 47,
188
+ 95,
189
+ 191,
190
+ 191,
191
+ 191
192
+ ]
193
  },
194
  "TS/M297.pt": {
 
 
 
195
  "dimension": "3",
196
+ "numberofchannels": "1",
197
+ "fov": [
198
+ 5,
199
+ 11,
200
+ 23,
201
+ 47,
202
+ 95,
203
+ 191,
204
+ 191
205
+ ]
206
  },
207
  "TS/M298.pt": {
 
 
 
208
  "dimension": "3",
209
+ "numberofchannels": "1",
210
+ "fov": [
211
+ 5,
212
+ 11,
213
+ 23,
214
+ 47,
215
+ 95,
216
+ 191,
217
+ 191
218
+ ]
219
  },
220
  "TS/M730.pt": {
 
 
 
221
  "dimension": "3",
222
+ "numberofchannels": "1",
223
+ "fov": [
224
+ 5,
225
+ 11,
226
+ 23,
227
+ 47,
228
+ 95,
229
+ 191,
230
+ 191,
231
+ 191
232
+ ]
233
  },
234
  "TS/M731.pt": {
 
 
 
235
  "dimension": "3",
236
+ "numberofchannels": "1",
237
+ "fov": [
238
+ 5,
239
+ 11,
240
+ 23,
241
+ 47,
242
+ 95,
243
+ 191,
244
+ 191,
245
+ 191
246
+ ]
247
  },
248
  "TS/M732.pt": {
 
 
 
249
  "dimension": "3",
250
+ "numberofchannels": "1",
251
+ "fov": [
252
+ 5,
253
+ 11,
254
+ 23,
255
+ 47,
256
+ 95,
257
+ 191,
258
+ 191
259
+ ]
260
  },
261
  "TS/M733.pt": {
 
 
 
262
  "dimension": "3",
263
+ "numberofchannels": "1",
264
+ "fov": [
265
+ 5,
266
+ 11,
267
+ 23,
268
+ 47,
269
+ 95,
270
+ 191
271
+ ]
272
  },
273
  "TS/M850.pt": {
 
 
 
274
  "dimension": "3",
275
+ "numberofchannels": "1",
276
+ "fov": [
277
+ 5,
278
+ 11,
279
+ 23,
280
+ 47,
281
+ 95,
282
+ 191,
283
+ 191,
284
+ 191
285
+ ]
286
  },
287
  "TS/M851.pt": {
 
 
 
288
  "dimension": "3",
289
+ "numberofchannels": "1",
290
+ "fov": [
291
+ 5,
292
+ 11,
293
+ 23,
294
+ 47,
295
+ 95,
296
+ 191,
297
+ 191,
298
+ 191
299
+ ]
300
  },
301
  "TS/M852.pt": {
 
 
 
302
  "dimension": "3",
303
+ "numberofchannels": "1",
304
+ "fov": [
305
+ 5,
306
+ 11,
307
+ 23,
308
+ 47,
309
+ 95,
310
+ 191,
311
+ 191
312
+ ]
313
  },
314
  "TS/M853.pt": {
 
 
 
315
  "dimension": "3",
316
+ "numberofchannels": "1",
317
+ "fov": [
318
+ 5,
319
+ 11,
320
+ 23,
321
+ 47,
322
+ 95,
323
+ 191
324
+ ]
325
  },
326
  "VGG/VGG16.pt": {
 
 
 
327
  "dimension": "2",
328
+ "numberofchannels": "3",
329
+ "fov": [
330
+ 5,
331
+ 11,
332
+ 23,
333
+ 47,
334
+ 95
335
+ ]
336
  }
337
  }