Hugging Face's logo

1aurent
/
vit_base_patch16_224.owkin_pancancer

Feature Extraction
timm
PyTorch
Safetensors
image-classification
biology
cancer
owkin
histology
Eval Results
Model card Files Community
1aurent commited on
Commit
0781fda
•
1 Parent(s): 6957904

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +165 -0
README.md CHANGED
@@ -8,6 +8,163 @@ tags:
8
  - owkin
9
  - histology
10
  library_name: timm
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
11
  widget:
12
  - src: https://github.com/owkin/HistoSSLscaling/raw/main/assets/example.tif
13
  example_title: pancancer tile
@@ -20,6 +177,12 @@ co2_eq_emissions:
20
  license: other
21
  pipeline_tag: feature-extraction
22
  inference: false
 
 
 
 
 
 
23
  ---
24
 
25
  # Model card for vit_base_patch16_224.owkin_pancancer
@@ -27,6 +190,8 @@ inference: false
27
  A Vision Transformer (ViT) image classification model. \
28
  Trained by Owkin on 40M pan-cancer histology tiles from TCGA.
29
 
 
 
30
  ## Model Details
31
 
32
  - **Model Type:** Image classification / feature backbone
 
8
  - owkin
9
  - histology
10
  library_name: timm
11
+ model-index:
12
+ - name: owkin_pancancer
13
+ results:
14
+ - task:
15
+ type: image-classification
16
+ name: Image Classification
17
+ dataset:
18
+ name: Camelyon16[Meta]
19
+ type: image-classification
20
+ metrics:
21
+ - type: accuracy
22
+ value: 94.5 ± 4.4
23
+ name: ROC AUC
24
+ verified: false
25
+ - task:
26
+ type: image-classification
27
+ name: Image Classification
28
+ dataset:
29
+ name: TCGA-BRCA[Hist]
30
+ type: image-classification
31
+ metrics:
32
+ - type: accuracy
33
+ value: 96.2 ± 3.3
34
+ name: ROC AUC
35
+ verified: false
36
+ - task:
37
+ type: image-classification
38
+ name: Image Classification
39
+ dataset:
40
+ name: TCGA-BRCA[HRD]
41
+ type: image-classification
42
+ metrics:
43
+ - type: accuracy
44
+ value: 79.3 ± 2.4
45
+ name: ROC AUC
46
+ verified: false
47
+ - task:
48
+ type: image-classification
49
+ name: Image Classification
50
+ dataset:
51
+ name: TCGA-BRCA[Mol]
52
+ type: image-classification
53
+ metrics:
54
+ - type: accuracy
55
+ value: 81.7 ± 1.6
56
+ name: ROC AUC
57
+ verified: false
58
+ - task:
59
+ type: image-classification
60
+ name: Image Classification
61
+ dataset:
62
+ name: TCGA-BRCA[OS]
63
+ type: image-classification
64
+ metrics:
65
+ - type: accuracy
66
+ value: 64.7 ± 5.7
67
+ name: ROC AUC
68
+ verified: false
69
+ - task:
70
+ type: image-classification
71
+ name: Image Classification
72
+ dataset:
73
+ name: TCGA-CRC[MSI]
74
+ type: image-classification
75
+ metrics:
76
+ - type: accuracy
77
+ value: 91.0 ± 2.2
78
+ name: ROC AUC
79
+ verified: false
80
+ - task:
81
+ type: image-classification
82
+ name: Image Classification
83
+ dataset:
84
+ name: TCGA-COAD[OS]
85
+ type: image-classification
86
+ metrics:
87
+ - type: accuracy
88
+ value: 63.4 ± 7.4
89
+ name: ROC AUC
90
+ verified: false
91
+ - task:
92
+ type: image-classification
93
+ name: Image Classification
94
+ dataset:
95
+ name: TCGA-NSCLC[CType]
96
+ type: image-classification
97
+ metrics:
98
+ - type: accuracy
99
+ value: 97.7 ± 1.3
100
+ name: ROC AUC
101
+ verified: false
102
+ - task:
103
+ type: image-classification
104
+ name: Image Classification
105
+ dataset:
106
+ name: TCGA-LUAD[OS]
107
+ type: image-classification
108
+ metrics:
109
+ - type: accuracy
110
+ value: 53.8 ± 4.5
111
+ name: ROC AUC
112
+ verified: false
113
+ - task:
114
+ type: image-classification
115
+ name: Image Classification
116
+ dataset:
117
+ name: TCGA-LUSC[OS]
118
+ type: image-classification
119
+ metrics:
120
+ - type: accuracy
121
+ value: 62.2 ± 2.9
122
+ name: ROC AUC
123
+ verified: false
124
+ - task:
125
+ type: image-classification
126
+ name: Image Classification
127
+ dataset:
128
+ name: TCGA-OV[HRD]
129
+ type: image-classification
130
+ metrics:
131
+ - type: accuracy
132
+ value: 74.2± 8.6
133
+ name: ROC AUC
134
+ verified: false
135
+ - task:
136
+ type: image-classification
137
+ name: Image Classification
138
+ dataset:
139
+ name: TCGA-RCC[CType]
140
+ type: image-classification
141
+ metrics:
142
+ - type: accuracy
143
+ value: 99.5 ± 0.2
144
+ name: ROC AUC
145
+ verified: false
146
+ - task:
147
+ type: image-classification
148
+ name: Image Classification
149
+ dataset:
150
+ name: TCGA-STAD[MSI]
151
+ type: image-classification
152
+ metrics:
153
+ - type: accuracy
154
+ value: 89.9 ± 3.9
155
+ name: ROC AUC
156
+ verified: false
157
+ - task:
158
+ type: image-classification
159
+ name: Image Classification
160
+ dataset:
161
+ name: TCGA-PAAD[OS]
162
+ type: image-classification
163
+ metrics:
164
+ - type: accuracy
165
+ value: 59.2 ± 4.1
166
+ name: ROC AUC
167
+ verified: false
168
  widget:
169
  - src: https://github.com/owkin/HistoSSLscaling/raw/main/assets/example.tif
170
  example_title: pancancer tile
 
177
  license: other
178
  pipeline_tag: feature-extraction
179
  inference: false
180
+ datasets:
181
+ - owkin/camelyon16-features
182
+ - owkin/nct-crc-he
183
+ - 1aurent/NCT-CRC-HE
184
+ metrics:
185
+ - roc_auc
186
  ---
187
 
188
  # Model card for vit_base_patch16_224.owkin_pancancer
 
190
  A Vision Transformer (ViT) image classification model. \
191
  Trained by Owkin on 40M pan-cancer histology tiles from TCGA.
192
 
193
+ ![](https://github.com/owkin/HistoSSLscaling/blob/main/assets/main_figure.png?raw=true)
194
+
195
  ## Model Details
196
 
197
  - **Model Type:** Image classification / feature backbone