Hugging Face's logo

sharmajai901
/
Watermark-detection

Image Classification
Transformers
English
siglip
Image-Classification
Watermark-Detection
SigLIP2
Model card Files
xet
Community
Jai Sharma commited on
Commit
989d134
·
1 Parent(s): c558b49

changes

Browse files
Files changed (16) hide show
  1. README.md +170 -0
  2. checkpoint-1424/config.json +46 -0
  3. checkpoint-1424/preprocessor_config.json +24 -0
  4. checkpoint-1424/rng_state.pth +3 -0
  5. checkpoint-1424/scheduler.pt +3 -0
  6. checkpoint-1424/trainer_state.json +68 -0
  7. checkpoint-1424/training_args.bin +3 -0
  8. checkpoint-712/config.json +46 -0
  9. checkpoint-712/preprocessor_config.json +24 -0
  10. checkpoint-712/rng_state.pth +3 -0
  11. checkpoint-712/scheduler.pt +3 -0
  12. checkpoint-712/trainer_state.json +51 -0
  13. checkpoint-712/training_args.bin +3 -0
  14. config.json +46 -0
  15. preprocessor_config.json +24 -0
  16. training_args.bin +3 -0
README.md ADDED
@@ -0,0 +1,170 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ datasets:
4
+ - qwertyforce/scenery_watermarks
5
+ language:
6
+ - en
7
+ base_model:
8
+ - google/siglip2-base-patch16-224
9
+ pipeline_tag: image-classification
10
+ library_name: transformers
11
+ tags:
12
+ - Image-Classification
13
+ - Watermark-Detection
14
+ - SigLIP2
15
+
16
+ ---
17
+
18
+ ![5.png](https://cdn-uploads.huggingface.co/production/uploads/65bb837dbfb878f46c77de4c/VXSOLkmcLM1t6XhTcYXUh.png)
19
+
20
+ # **Watermark-Detection-SigLIP2**
21
+
22
+ > **Watermark-Detection-SigLIP2** is a vision-language encoder model fine-tuned from **google/siglip2-base-patch16-224** for **binary image classification**. It is trained to detect whether an image **contains a watermark or not**, using the **SiglipForImageClassification** architecture.
23
+
24
+ > [!note]
25
+ > Watermark detection works best with crisp and high-quality images. Noisy images are not recommended for validation.
26
+
27
+ > [!note]
28
+ *SigLIP 2: Multilingual Vision-Language Encoders with Improved Semantic Understanding, Localization, and Dense Features* https://arxiv.org/pdf/2502.14786
29
+
30
+ ```py
31
+ Classification Report:
32
+ precision recall f1-score support
33
+
34
+ No Watermark 0.9290 0.9722 0.9501 12779
35
+ Watermark 0.9622 0.9048 0.9326 9983
36
+
37
+ accuracy 0.9427 22762
38
+ macro avg 0.9456 0.9385 0.9414 22762
39
+ weighted avg 0.9435 0.9427 0.9424 22762
40
+ ```
41
+
42
+ ![download.png](https://cdn-uploads.huggingface.co/production/uploads/65bb837dbfb878f46c77de4c/_rKqtbSJbglsRiXmRF1ij.png)
43
+
44
+ ---
45
+
46
+ ## **Label Space: 2 Classes**
47
+
48
+ The model classifies an image as either:
49
+
50
+ ```
51
+ Class 0: "No Watermark"
52
+ Class 1: "Watermark"
53
+ ```
54
+
55
+ ---
56
+
57
+ ## **Install dependencies**
58
+
59
+ ```bash
60
+ pip install -q transformers torch pillow gradio
61
+ ```
62
+
63
+ ---
64
+
65
+ ## **Inference Code**
66
+
67
+ ```python
68
+ import gradio as gr
69
+ from transformers import AutoImageProcessor, SiglipForImageClassification
70
+ from PIL import Image
71
+ import torch
72
+
73
+ # Load model and processor
74
+ model_name = "prithivMLmods/Watermark-Detection-SigLIP2" # Update this if using a different path
75
+ model = SiglipForImageClassification.from_pretrained(model_name)
76
+ processor = AutoImageProcessor.from_pretrained(model_name)
77
+
78
+ # Label mapping
79
+ id2label = {
80
+ "0": "No Watermark",
81
+ "1": "Watermark"
82
+ }
83
+
84
+ def classify_watermark(image):
85
+ image = Image.fromarray(image).convert("RGB")
86
+ inputs = processor(images=image, return_tensors="pt")
87
+
88
+ with torch.no_grad():
89
+ outputs = model(**inputs)
90
+ logits = outputs.logits
91
+ probs = torch.nn.functional.softmax(logits, dim=1).squeeze().tolist()
92
+
93
+ prediction = {
94
+ id2label[str(i)]: round(probs[i], 3) for i in range(len(probs))
95
+ }
96
+
97
+ return prediction
98
+
99
+ # Gradio Interface
100
+ iface = gr.Interface(
101
+ fn=classify_watermark,
102
+ inputs=gr.Image(type="numpy"),
103
+ outputs=gr.Label(num_top_classes=2, label="Watermark Detection"),
104
+ title="Watermark-Detection-SigLIP2",
105
+ description="Upload an image to detect whether it contains a watermark."
106
+ )
107
+
108
+ if __name__ == "__main__":
109
+ iface.launch()
110
+ ```
111
+
112
+ ---
113
+
114
+ ## **Demo Inference**
115
+
116
+ > [!Warning]
117
+ > Watermark
118
+
119
+ <table>
120
+ <tr>
121
+ <td><img src="https://cdn-uploads.huggingface.co/production/uploads/65bb837dbfb878f46c77de4c/sm062kFE7QJiLisTTjNwv.png" width="300"/></td>
122
+ <td><img src="https://cdn-uploads.huggingface.co/production/uploads/65bb837dbfb878f46c77de4c/UFymm_tzVRmov6vn_cElE.png" width="300"/></td>
123
+ <td><img src="https://cdn-uploads.huggingface.co/production/uploads/65bb837dbfb878f46c77de4c/bPzPAK-Mib8nFhHCkjD2B.png" width="300"/></td>
124
+ </tr>
125
+ <tr>
126
+ <td><img src="https://cdn-uploads.huggingface.co/production/uploads/65bb837dbfb878f46c77de4c/4fP8SBIYofKEeDBU0klQ2.png" width="300"/></td>
127
+ <td><img src="https://cdn-uploads.huggingface.co/production/uploads/65bb837dbfb878f46c77de4c/wD5M4YgyQGk9-QLFjMcn9.png" width="300"/></td>
128
+ <td><img src="https://cdn-uploads.huggingface.co/production/uploads/65bb837dbfb878f46c77de4c/yg0q88-0S4k4FUS4-qGNw.png" width="300"/></td>
129
+ </tr>
130
+ <tr>
131
+ <td><img src="https://cdn-uploads.huggingface.co/production/uploads/65bb837dbfb878f46c77de4c/WhRkeYw8-wIgldpaz0E4m.png" width="300"/></td>
132
+ <td><img src="https://cdn-uploads.huggingface.co/production/uploads/65bb837dbfb878f46c77de4c/Uhb1zBxQV_5CWLoyTAMmD.png" width="300"/></td>
133
+ <td><img src="https://cdn-uploads.huggingface.co/production/uploads/65bb837dbfb878f46c77de4c/7hnLD2b0f7B7edwgx_eOR.png" width="300"/></td>
134
+ </tr>
135
+ </table>
136
+
137
+ > [!Warning]
138
+ > No Watermark
139
+
140
+ <table>
141
+ <tr>
142
+ <td><img src="https://cdn-uploads.huggingface.co/production/uploads/65bb837dbfb878f46c77de4c/edyFBIETs3Dosn1edpGZ8.png" width="300"/></td>
143
+ <td><img src="https://cdn-uploads.huggingface.co/production/uploads/65bb837dbfb878f46c77de4c/3bRMcr2r0k00mMkthbYDW.png" width="300"/></td>
144
+ <td><img src="https://cdn-uploads.huggingface.co/production/uploads/65bb837dbfb878f46c77de4c/eeMLQEg4r89f9owe8jSij.png" width="300"/></td>
145
+ </tr>
146
+ <tr>
147
+ <td><img src="https://cdn-uploads.huggingface.co/production/uploads/65bb837dbfb878f46c77de4c/45jk4dvZk1wT3L7cprqql.png" width="300"/></td>
148
+ <td><img src="https://cdn-uploads.huggingface.co/production/uploads/65bb837dbfb878f46c77de4c/mrkm0JXXgSQVXi0_d7EKH.png" width="300"/></td>
149
+ <td><img src="https://cdn-uploads.huggingface.co/production/uploads/65bb837dbfb878f46c77de4c/f_5R7Inb8I-32hWJchkgj.png" width="300"/></td>
150
+ </tr>
151
+ <tr>
152
+ <td><img src="https://cdn-uploads.huggingface.co/production/uploads/65bb837dbfb878f46c77de4c/qIUTSy8SuJEsRkYGd0L5d.png" width="300"/></td>
153
+ <td><img src="https://cdn-uploads.huggingface.co/production/uploads/65bb837dbfb878f46c77de4c/DnlNo9lM4mBNUjlexKLVa.png" width="300"/></td>
154
+ <td><img src="https://cdn-uploads.huggingface.co/production/uploads/65bb837dbfb878f46c77de4c/bs4oyaapW8mi0lizOqWSf.png" width="300"/></td>
155
+ </tr>
156
+ </table>
157
+
158
+ ---
159
+
160
+ ## **Intended Use**
161
+
162
+ **Watermark-Detection-SigLIP2** is useful in scenarios such as:
163
+
164
+ - **Content Moderation** – Automatically detect watermarked content on image sharing platforms.
165
+ - **Dataset Cleaning** – Filter out watermarked images from training datasets.
166
+ - **Copyright Enforcement** – Monitor and flag usage of watermarked media.
167
+ - **Digital Forensics** – Support analysis of tampered or protected media assets.
168
+
169
+
170
+
checkpoint-1424/config.json ADDED
@@ -0,0 +1,46 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "SiglipForImageClassification"
4
+ ],
5
+ "id2label": {
6
+ "0": "No Watermark",
7
+ "1": "Watermark"
8
+ },
9
+ "initializer_factor": 1.0,
10
+ "label2id": {
11
+ "No Watermark": 0,
12
+ "Watermark": 1
13
+ },
14
+ "model_type": "siglip",
15
+ "problem_type": "single_label_classification",
16
+ "text_config": {
17
+ "attention_dropout": 0.0,
18
+ "hidden_act": "gelu_pytorch_tanh",
19
+ "hidden_size": 768,
20
+ "intermediate_size": 3072,
21
+ "layer_norm_eps": 1e-06,
22
+ "max_position_embeddings": 64,
23
+ "model_type": "siglip_text_model",
24
+ "num_attention_heads": 12,
25
+ "num_hidden_layers": 12,
26
+ "projection_size": 768,
27
+ "torch_dtype": "float32",
28
+ "vocab_size": 256000
29
+ },
30
+ "torch_dtype": "float32",
31
+ "transformers_version": "4.50.0",
32
+ "vision_config": {
33
+ "attention_dropout": 0.0,
34
+ "hidden_act": "gelu_pytorch_tanh",
35
+ "hidden_size": 768,
36
+ "image_size": 224,
37
+ "intermediate_size": 3072,
38
+ "layer_norm_eps": 1e-06,
39
+ "model_type": "siglip_vision_model",
40
+ "num_attention_heads": 12,
41
+ "num_channels": 3,
42
+ "num_hidden_layers": 12,
43
+ "patch_size": 16,
44
+ "torch_dtype": "float32"
45
+ }
46
+ }
checkpoint-1424/preprocessor_config.json ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "do_convert_rgb": null,
3
+ "do_normalize": true,
4
+ "do_rescale": true,
5
+ "do_resize": true,
6
+ "image_mean": [
7
+ 0.5,
8
+ 0.5,
9
+ 0.5
10
+ ],
11
+ "image_processor_type": "SiglipImageProcessor",
12
+ "image_std": [
13
+ 0.5,
14
+ 0.5,
15
+ 0.5
16
+ ],
17
+ "processor_class": "SiglipProcessor",
18
+ "resample": 2,
19
+ "rescale_factor": 0.00392156862745098,
20
+ "size": {
21
+ "height": 224,
22
+ "width": 224
23
+ }
24
+ }
checkpoint-1424/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:13b2b1a200d366a090241026c516010359252c9ecfd83705f10861a52b5c39e9
3
+ size 14244
checkpoint-1424/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:847c98626d39d680c32b284d7139041edc61ef79d4947f1060da6b500567b4aa
3
+ size 1064
checkpoint-1424/trainer_state.json ADDED
@@ -0,0 +1,68 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": 1424,
3
+ "best_metric": 0.23910000920295715,
4
+ "best_model_checkpoint": "siglip2-finetune-full/checkpoint-1424",
5
+ "epoch": 2.0,
6
+ "eval_steps": 500,
7
+ "global_step": 1424,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.702247191011236,
14
+ "grad_norm": 16.442720413208008,
15
+ "learning_rate": 1.3449781659388646e-06,
16
+ "loss": 0.6162,
17
+ "step": 500
18
+ },
19
+ {
20
+ "epoch": 1.0,
21
+ "eval_accuracy": 0.9363852034091907,
22
+ "eval_loss": 0.2713288962841034,
23
+ "eval_model_preparation_time": 0.0023,
24
+ "eval_runtime": 335.7294,
25
+ "eval_samples_per_second": 67.799,
26
+ "eval_steps_per_second": 8.477,
27
+ "step": 712
28
+ },
29
+ {
30
+ "epoch": 1.404494382022472,
31
+ "grad_norm": 16.62710189819336,
32
+ "learning_rate": 6.171761280931587e-07,
33
+ "loss": 0.519,
34
+ "step": 1000
35
+ },
36
+ {
37
+ "epoch": 2.0,
38
+ "eval_accuracy": 0.9426676039012389,
39
+ "eval_loss": 0.23910000920295715,
40
+ "eval_model_preparation_time": 0.0023,
41
+ "eval_runtime": 338.2117,
42
+ "eval_samples_per_second": 67.301,
43
+ "eval_steps_per_second": 8.415,
44
+ "step": 1424
45
+ }
46
+ ],
47
+ "logging_steps": 500,
48
+ "max_steps": 1424,
49
+ "num_input_tokens_seen": 0,
50
+ "num_train_epochs": 2,
51
+ "save_steps": 500,
52
+ "stateful_callbacks": {
53
+ "TrainerControl": {
54
+ "args": {
55
+ "should_epoch_stop": false,
56
+ "should_evaluate": false,
57
+ "should_log": false,
58
+ "should_save": true,
59
+ "should_training_stop": true
60
+ },
61
+ "attributes": {}
62
+ }
63
+ },
64
+ "total_flos": 3.8128851917694075e+18,
65
+ "train_batch_size": 32,
66
+ "trial_name": null,
67
+ "trial_params": null
68
+ }
checkpoint-1424/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4138dcb234dcef0352a6fbae254152fc6e7544045ab1dbc0e451ec4366da1634
3
+ size 5304
checkpoint-712/config.json ADDED
@@ -0,0 +1,46 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "SiglipForImageClassification"
4
+ ],
5
+ "id2label": {
6
+ "0": "No Watermark",
7
+ "1": "Watermark"
8
+ },
9
+ "initializer_factor": 1.0,
10
+ "label2id": {
11
+ "No Watermark": 0,
12
+ "Watermark": 1
13
+ },
14
+ "model_type": "siglip",
15
+ "problem_type": "single_label_classification",
16
+ "text_config": {
17
+ "attention_dropout": 0.0,
18
+ "hidden_act": "gelu_pytorch_tanh",
19
+ "hidden_size": 768,
20
+ "intermediate_size": 3072,
21
+ "layer_norm_eps": 1e-06,
22
+ "max_position_embeddings": 64,
23
+ "model_type": "siglip_text_model",
24
+ "num_attention_heads": 12,
25
+ "num_hidden_layers": 12,
26
+ "projection_size": 768,
27
+ "torch_dtype": "float32",
28
+ "vocab_size": 256000
29
+ },
30
+ "torch_dtype": "float32",
31
+ "transformers_version": "4.50.0",
32
+ "vision_config": {
33
+ "attention_dropout": 0.0,
34
+ "hidden_act": "gelu_pytorch_tanh",
35
+ "hidden_size": 768,
36
+ "image_size": 224,
37
+ "intermediate_size": 3072,
38
+ "layer_norm_eps": 1e-06,
39
+ "model_type": "siglip_vision_model",
40
+ "num_attention_heads": 12,
41
+ "num_channels": 3,
42
+ "num_hidden_layers": 12,
43
+ "patch_size": 16,
44
+ "torch_dtype": "float32"
45
+ }
46
+ }
checkpoint-712/preprocessor_config.json ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "do_convert_rgb": null,
3
+ "do_normalize": true,
4
+ "do_rescale": true,
5
+ "do_resize": true,
6
+ "image_mean": [
7
+ 0.5,
8
+ 0.5,
9
+ 0.5
10
+ ],
11
+ "image_processor_type": "SiglipImageProcessor",
12
+ "image_std": [
13
+ 0.5,
14
+ 0.5,
15
+ 0.5
16
+ ],
17
+ "processor_class": "SiglipProcessor",
18
+ "resample": 2,
19
+ "rescale_factor": 0.00392156862745098,
20
+ "size": {
21
+ "height": 224,
22
+ "width": 224
23
+ }
24
+ }
checkpoint-712/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:57ca47b1e99fcaf46bc056bcb92bed3455a48faa74111373f1b003d76595409e
3
+ size 14244
checkpoint-712/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fa080b4dbab1f138b2cecf9d7402c5cbaaa81390a9edbd3ef8c30ae273047153
3
+ size 1064
checkpoint-712/trainer_state.json ADDED
@@ -0,0 +1,51 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": 712,
3
+ "best_metric": 0.2713288962841034,
4
+ "best_model_checkpoint": "siglip2-finetune-full/checkpoint-712",
5
+ "epoch": 1.0,
6
+ "eval_steps": 500,
7
+ "global_step": 712,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.702247191011236,
14
+ "grad_norm": 16.442720413208008,
15
+ "learning_rate": 1.3449781659388646e-06,
16
+ "loss": 0.6162,
17
+ "step": 500
18
+ },
19
+ {
20
+ "epoch": 1.0,
21
+ "eval_accuracy": 0.9363852034091907,
22
+ "eval_loss": 0.2713288962841034,
23
+ "eval_model_preparation_time": 0.0023,
24
+ "eval_runtime": 335.7294,
25
+ "eval_samples_per_second": 67.799,
26
+ "eval_steps_per_second": 8.477,
27
+ "step": 712
28
+ }
29
+ ],
30
+ "logging_steps": 500,
31
+ "max_steps": 1424,
32
+ "num_input_tokens_seen": 0,
33
+ "num_train_epochs": 2,
34
+ "save_steps": 500,
35
+ "stateful_callbacks": {
36
+ "TrainerControl": {
37
+ "args": {
38
+ "should_epoch_stop": false,
39
+ "should_evaluate": false,
40
+ "should_log": false,
41
+ "should_save": true,
42
+ "should_training_stop": false
43
+ },
44
+ "attributes": {}
45
+ }
46
+ },
47
+ "total_flos": 1.9064425958847037e+18,
48
+ "train_batch_size": 32,
49
+ "trial_name": null,
50
+ "trial_params": null
51
+ }
checkpoint-712/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4138dcb234dcef0352a6fbae254152fc6e7544045ab1dbc0e451ec4366da1634
3
+ size 5304
config.json ADDED
@@ -0,0 +1,46 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "SiglipForImageClassification"
4
+ ],
5
+ "id2label": {
6
+ "0": "No Watermark",
7
+ "1": "Watermark"
8
+ },
9
+ "initializer_factor": 1.0,
10
+ "label2id": {
11
+ "No Watermark": 0,
12
+ "Watermark": 1
13
+ },
14
+ "model_type": "siglip",
15
+ "problem_type": "single_label_classification",
16
+ "text_config": {
17
+ "attention_dropout": 0.0,
18
+ "hidden_act": "gelu_pytorch_tanh",
19
+ "hidden_size": 768,
20
+ "intermediate_size": 3072,
21
+ "layer_norm_eps": 1e-06,
22
+ "max_position_embeddings": 64,
23
+ "model_type": "siglip_text_model",
24
+ "num_attention_heads": 12,
25
+ "num_hidden_layers": 12,
26
+ "projection_size": 768,
27
+ "torch_dtype": "float32",
28
+ "vocab_size": 256000
29
+ },
30
+ "torch_dtype": "float32",
31
+ "transformers_version": "4.50.0",
32
+ "vision_config": {
33
+ "attention_dropout": 0.0,
34
+ "hidden_act": "gelu_pytorch_tanh",
35
+ "hidden_size": 768,
36
+ "image_size": 224,
37
+ "intermediate_size": 3072,
38
+ "layer_norm_eps": 1e-06,
39
+ "model_type": "siglip_vision_model",
40
+ "num_attention_heads": 12,
41
+ "num_channels": 3,
42
+ "num_hidden_layers": 12,
43
+ "patch_size": 16,
44
+ "torch_dtype": "float32"
45
+ }
46
+ }
preprocessor_config.json ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "do_convert_rgb": null,
3
+ "do_normalize": true,
4
+ "do_rescale": true,
5
+ "do_resize": true,
6
+ "image_mean": [
7
+ 0.5,
8
+ 0.5,
9
+ 0.5
10
+ ],
11
+ "image_processor_type": "SiglipImageProcessor",
12
+ "image_std": [
13
+ 0.5,
14
+ 0.5,
15
+ 0.5
16
+ ],
17
+ "processor_class": "SiglipProcessor",
18
+ "resample": 2,
19
+ "rescale_factor": 0.00392156862745098,
20
+ "size": {
21
+ "height": 224,
22
+ "width": 224
23
+ }
24
+ }
training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4138dcb234dcef0352a6fbae254152fc6e7544045ab1dbc0e451ec4366da1634
3
+ size 5304