Hugging Face's logo

pabloOmega
/
ocr-entities-detection-ft

Image Segmentation
Transformers
Safetensors
detr
Model card Files
xet
Community
pabloOmega commited on
Commit
ae1b80d
·
verified ·
1 Parent(s): 1e903c5

Upload DetrForSegmentation

Browse files
Files changed (2) hide show
  1. config.json +62 -41
  2. model.safetensors +2 -2
config.json CHANGED
@@ -1,59 +1,80 @@
1
  {
 
 
2
  "architectures": [
3
- "YolosForObjectDetection"
4
  ],
5
- "attention_probs_dropout_prob": 0.0,
6
  "auxiliary_loss": false,
 
 
 
 
 
 
 
 
 
 
 
7
  "bbox_cost": 5,
8
  "bbox_loss_coefficient": 5,
9
  "class_cost": 1,
 
 
 
 
 
 
 
 
 
 
 
 
 
10
  "eos_coefficient": 0.1,
11
  "giou_cost": 2,
12
  "giou_loss_coefficient": 2,
13
- "hidden_act": "gelu",
14
- "hidden_dropout_prob": 0.0,
15
- "hidden_size": 384,
16
  "id2label": {
17
- "0": "LABEL_0",
18
- "1": "LABEL_1",
19
- "2": "LABEL_2",
20
- "3": "LABEL_3",
21
- "4": "LABEL_4",
22
- "5": "LABEL_5",
23
- "6": "LABEL_6",
24
- "7": "LABEL_7",
25
- "8": "LABEL_8",
26
- "9": "LABEL_9",
27
- "10": "LABEL_10"
28
  },
29
- "image_size": [
30
- 512,
31
- 864
32
- ],
33
- "initializer_range": 0.02,
34
- "intermediate_size": 1536,
35
  "label2id": {
36
- "LABEL_0": 0,
37
- "LABEL_1": 1,
38
- "LABEL_10": 10,
39
- "LABEL_2": 2,
40
- "LABEL_3": 3,
41
- "LABEL_4": 4,
42
- "LABEL_5": 5,
43
- "LABEL_6": 6,
44
- "LABEL_7": 7,
45
- "LABEL_8": 8,
46
- "LABEL_9": 9
47
  },
48
- "layer_norm_eps": 1e-12,
49
- "model_type": "yolos",
50
- "num_attention_heads": 6,
51
  "num_channels": 3,
52
- "num_detection_tokens": 100,
53
- "num_hidden_layers": 12,
54
- "patch_size": 16,
55
- "qkv_bias": true,
56
  "torch_dtype": "float32",
57
  "transformers_version": "4.50.3",
58
- "use_mid_position_embeddings": true
 
59
  }
 
1
  {
2
+ "activation_dropout": 0.0,
3
+ "activation_function": "relu",
4
  "architectures": [
5
+ "DetrForSegmentation"
6
  ],
7
+ "attention_dropout": 0.0,
8
  "auxiliary_loss": false,
9
+ "backbone": "resnet50",
10
+ "backbone_config": null,
11
+ "backbone_kwargs": {
12
+ "in_chans": 3,
13
+ "out_indices": [
14
+ 1,
15
+ 2,
16
+ 3,
17
+ 4
18
+ ]
19
+ },
20
  "bbox_cost": 5,
21
  "bbox_loss_coefficient": 5,
22
  "class_cost": 1,
23
+ "classifier_dropout": 0.0,
24
+ "d_model": 256,
25
+ "decoder_attention_heads": 8,
26
+ "decoder_ffn_dim": 2048,
27
+ "decoder_layerdrop": 0.0,
28
+ "decoder_layers": 6,
29
+ "dice_loss_coefficient": 1,
30
+ "dilation": false,
31
+ "dropout": 0.1,
32
+ "encoder_attention_heads": 8,
33
+ "encoder_ffn_dim": 2048,
34
+ "encoder_layerdrop": 0.0,
35
+ "encoder_layers": 6,
36
  "eos_coefficient": 0.1,
37
  "giou_cost": 2,
38
  "giou_loss_coefficient": 2,
 
 
 
39
  "id2label": {
40
+ "0": "Caption",
41
+ "1": "Footnote",
42
+ "2": "Formula",
43
+ "3": "List-item",
44
+ "4": "Page-footer",
45
+ "5": "Page-header",
46
+ "6": "Picture",
47
+ "7": "Section-header",
48
+ "8": "Table",
49
+ "9": "Text",
50
+ "10": "Title"
51
  },
52
+ "init_std": 0.02,
53
+ "init_xavier_std": 1.0,
54
+ "is_encoder_decoder": true,
 
 
 
55
  "label2id": {
56
+ "Caption": 0,
57
+ "Footnote": 1,
58
+ "Formula": 2,
59
+ "List-item": 3,
60
+ "Page-footer": 4,
61
+ "Page-header": 5,
62
+ "Picture": 6,
63
+ "Section-header": 7,
64
+ "Table": 8,
65
+ "Text": 9,
66
+ "Title": 10
67
  },
68
+ "mask_loss_coefficient": 1,
69
+ "max_position_embeddings": 1024,
70
+ "model_type": "detr",
71
  "num_channels": 3,
72
+ "num_hidden_layers": 6,
73
+ "num_queries": 100,
74
+ "position_embedding_type": "sine",
75
+ "scale_embedding": false,
76
  "torch_dtype": "float32",
77
  "transformers_version": "4.50.3",
78
+ "use_pretrained_backbone": true,
79
+ "use_timm_backbone": true
80
  }
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:f803e8a62c89075babc27ca5a10ff557ad424b9387845bc2dde64712d7571412
3
- size 122640072
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f060f121030be409a597cbd0916ae3cc00a02db5953b784114f9f7aaf121df37
3
+ size 171845444