Hugging Face's logo

ClementP
/
fundus-lesions-segmentation-unet_seresnext50_32x4d

Transformers
Safetensors
English
Eval Results
Inference Endpoints
Model card Files Community
1
ClementP commited on
Commit
a0096ba
•
1 Parent(s): 8b9b4e6

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +122 -6
README.md CHANGED
@@ -1,9 +1,125 @@
1
  ---
2
- tags:
3
- - pytorch_model_hub_mixin
4
- - model_hub_mixin
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6
 
7
- This model has been pushed to the Hub using ****:
8
- - Repo: [More Information Needed]
9
- - Docs: [More Information Needed]
 
1
  ---
2
+ language: en
3
+ license: mit
4
+ datasets:
5
+ - DDR
6
+ - FGADR
7
+ - IDRID
8
+ - MESSIDOR
9
+ - RETLES
10
+ library: torchSeg
11
+ model-index:
12
+ - name: unet_seresnext50_32x4d
13
+ results:
14
+ - task:
15
+ type: image-segmentation
16
+ dataset:
17
+ name: IDRID
18
+ type: IDRID
19
+ metrics:
20
+ - type: roc_auc
21
+ value: 0.6701094508171082
22
+ name: AUC Precision Recall - IDRID COTTON_WOOL_SPOT - COTTON_WOOL_SPOT
23
+ - type: roc_auc
24
+ value: 0.7860875129699707
25
+ name: AUC Precision Recall - IDRID EXUDATES - EXUDATES
26
+ - type: roc_auc
27
+ value: 0.6743975877761841
28
+ name: AUC Precision Recall - IDRID HEMORRHAGES - HEMORRHAGES
29
+ - type: roc_auc
30
+ value: 0.39846163988113403
31
+ name: AUC Precision Recall - IDRID MICROANEURYSMS - MICROANEURYSMS
32
+ - task:
33
+ type: image-segmentation
34
+ dataset:
35
+ name: FGADR
36
+ type: FGADR
37
+ metrics:
38
+ - type: roc_auc
39
+ value: 0.4449217915534973
40
+ name: AUC Precision Recall - FGADR COTTON_WOOL_SPOT - COTTON_WOOL_SPOT
41
+ - type: roc_auc
42
+ value: 0.6951484084129333
43
+ name: AUC Precision Recall - FGADR EXUDATES - EXUDATES
44
+ - type: roc_auc
45
+ value: 0.6508341431617737
46
+ name: AUC Precision Recall - FGADR HEMORRHAGES - HEMORRHAGES
47
+ - type: roc_auc
48
+ value: 0.2895563244819641
49
+ name: AUC Precision Recall - FGADR MICROANEURYSMS - MICROANEURYSMS
50
+ - task:
51
+ type: image-segmentation
52
+ dataset:
53
+ name: MESSIDOR
54
+ type: MESSIDOR
55
+ metrics:
56
+ - type: roc_auc
57
+ value: 0.3307325839996338
58
+ name: AUC Precision Recall - MESSIDOR COTTON_WOOL_SPOT - COTTON_WOOL_SPOT
59
+ - type: roc_auc
60
+ value: 0.7123324871063232
61
+ name: AUC Precision Recall - MESSIDOR EXUDATES - EXUDATES
62
+ - type: roc_auc
63
+ value: 0.3926454186439514
64
+ name: AUC Precision Recall - MESSIDOR HEMORRHAGES - HEMORRHAGES
65
+ - type: roc_auc
66
+ value: 0.4098129868507385
67
+ name: AUC Precision Recall - MESSIDOR MICROANEURYSMS - MICROANEURYSMS
68
+ - task:
69
+ type: image-segmentation
70
+ dataset:
71
+ name: DDR
72
+ type: DDR
73
+ metrics:
74
+ - type: roc_auc
75
+ value: 0.5084977746009827
76
+ name: AUC Precision Recall - DDR COTTON_WOOL_SPOT - COTTON_WOOL_SPOT
77
+ - type: roc_auc
78
+ value: 0.6117375493049622
79
+ name: AUC Precision Recall - DDR EXUDATES - EXUDATES
80
+ - type: roc_auc
81
+ value: 0.5447860956192017
82
+ name: AUC Precision Recall - DDR HEMORRHAGES - HEMORRHAGES
83
+ - type: roc_auc
84
+ value: 0.23405438661575317
85
+ name: AUC Precision Recall - DDR MICROANEURYSMS - MICROANEURYSMS
86
+ - task:
87
+ type: image-segmentation
88
+ dataset:
89
+ name: RETLES
90
+ type: RETLES
91
+ metrics:
92
+ - type: roc_auc
93
+ value: 0.5254419445991516
94
+ name: AUC Precision Recall - RETLES COTTON_WOOL_SPOT - COTTON_WOOL_SPOT
95
+ - type: roc_auc
96
+ value: 0.7039055824279785
97
+ name: AUC Precision Recall - RETLES EXUDATES - EXUDATES
98
+ - type: roc_auc
99
+ value: 0.5196094512939453
100
+ name: AUC Precision Recall - RETLES HEMORRHAGES - HEMORRHAGES
101
+ - type: roc_auc
102
+ value: 0.4127877354621887
103
+ name: AUC Precision Recall - RETLES MICROANEURYSMS - MICROANEURYSMS
104
  ---
105
+ # Lesions Segmentation in Fundus
106
+
107
+ ## Introduction
108
+ We focus on the semantic segmentations of:
109
+
110
+ 1. Cotton Wool Spot
111
+ 2. Exudates
112
+ 3. Hemmorrhages
113
+ 4. Microaneurysms
114
+
115
+ For an easier use of the models, we refer to cleaned-up version of the code provided in the [fundus lesions toolkit](https://github.com/ClementPla/fundus-lesions-toolkit/tree/main/).
116
+
117
+ ## Architecture
118
+
119
+ The model uses unet_seresnext50_32x4d as architecture. The implementation is taken from [torchSeg](https://github.com/isaaccorley/torchseg)
120
+
121
+ ## Training datasets
122
+
123
+ The model was trained on the following datasets:
124
+ DDR, FGADR, IDRID, MESSIDOR, RETLES
125