Hugging Face's logo

VinayHajare
/
siglip2-finetuned-marathi-sign-language

Image Classification
Transformers
Safetensors
Marathi
siglip
Generated from Trainer
Model card Files Community
VinayHajare commited on
Commit
75a6148
·
verified ·
1 Parent(s): 05bc899

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +120 -68
README.md CHANGED
@@ -1,68 +1,120 @@
1
- ---
2
- library_name: transformers
3
- license: apache-2.0
4
- base_model: google/siglip2-base-patch16-224
5
- tags:
6
- - generated_from_trainer
7
- metrics:
8
- - accuracy
9
- model-index:
10
- - name: siglip2-finetuned-marathi-sign-language
11
- results: []
12
- ---
13
-
14
- <!-- This model card has been generated automatically according to the information the Trainer had access to. You
15
- should probably proofread and complete it, then remove this comment. -->
16
-
17
- # siglip2-finetuned-marathi-sign-language
18
-
19
- This model is a fine-tuned version of [google/siglip2-base-patch16-224](https://huggingface.co/google/siglip2-base-patch16-224) on an unknown dataset.
20
- It achieves the following results on the evaluation set:
21
- - Loss: 0.0006
22
- - Model Preparation Time: 0.0057
23
- - Accuracy: 0.9997
24
-
25
- ## Model description
26
-
27
- More information needed
28
-
29
- ## Intended uses & limitations
30
-
31
- More information needed
32
-
33
- ## Training and evaluation data
34
-
35
- More information needed
36
-
37
- ## Training procedure
38
-
39
- ### Training hyperparameters
40
-
41
- The following hyperparameters were used during training:
42
- - learning_rate: 2e-06
43
- - train_batch_size: 32
44
- - eval_batch_size: 8
45
- - seed: 42
46
- - optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
47
- - lr_scheduler_type: linear
48
- - lr_scheduler_warmup_steps: 50
49
- - num_epochs: 6
50
-
51
- ### Training results
52
-
53
- | Training Loss | Epoch | Step | Validation Loss | Model Preparation Time | Accuracy |
54
- |:-------------:|:-----:|:----:|:---------------:|:----------------------:|:--------:|
55
- | 1.4439 | 1.0 | 940 | 0.0090 | 0.0057 | 0.9980 |
56
- | 0.0052 | 2.0 | 1880 | 0.0035 | 0.0057 | 0.9993 |
57
- | 0.0031 | 3.0 | 2820 | 0.0016 | 0.0057 | 0.9997 |
58
- | 0.001 | 4.0 | 3760 | 0.0010 | 0.0057 | 0.9997 |
59
- | 0.0007 | 5.0 | 4700 | 0.0013 | 0.0057 | 0.9997 |
60
- | 0.0005 | 6.0 | 5640 | 0.0006 | 0.0057 | 0.9997 |
61
-
62
-
63
- ### Framework versions
64
-
65
- - Transformers 4.52.0.dev0
66
- - Pytorch 2.6.0+cu124
67
- - Datasets 3.5.1
68
- - Tokenizers 0.21.1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: transformers
3
+ license: apache-2.0
4
+ base_model: google/siglip2-base-patch16-224
5
+ tags:
6
+ - generated_from_trainer
7
+ metrics:
8
+ - accuracy
9
+ model-index:
10
+ - name: siglip2-finetuned-marathi-sign-language
11
+ results: []
12
+ datasets:
13
+ - VinayHajare/Marathi-Sign-Language
14
+ language:
15
+ - mr
16
+ pipeline_tag: image-classification
17
+ ---
18
+
19
+ <!-- This model card has been generated automatically according to the information the Trainer had access to. You
20
+ should probably proofread and complete it, then remove this comment. -->
21
+
22
+ # siglip2-finetuned-marathi-sign-language
23
+
24
+ This model is a fine-tuned version of [google/siglip2-base-patch16-224](https://huggingface.co/google/siglip2-base-patch16-224) on an unknown dataset.
25
+ It achieves the following results on the evaluation set:
26
+ - Loss: 0.0006
27
+ - Model Preparation Time: 0.0057
28
+ - Accuracy: 0.9997
29
+
30
+ ## Model description
31
+
32
+ Marathi-Sign-Language-Detection is a vision-language model fine-tuned from google/siglip2-base-patch16-224 for multi-class image classification. It is trained to recognize Marathi sign language hand gestures and map them to corresponding Devanagari characters using the SiglipForImageClassification architecture.
33
+
34
+ ## Training and evaluation data
35
+
36
+ ```java
37
+ Classification Report:
38
+ precision recall f1-score support
39
+
40
+ अ 1.0000 1.0000 1.0000 404
41
+ आ 1.0000 1.0000 1.0000 409
42
+ इ 1.0000 1.0000 1.0000 440
43
+ ई 0.9866 1.0000 0.9932 441
44
+ उ 1.0000 1.0000 1.0000 479
45
+ ऊ 1.0000 1.0000 1.0000 428
46
+ ए 1.0000 1.0000 1.0000 457
47
+ ऐ 1.0000 1.0000 1.0000 436
48
+ ओ 1.0000 1.0000 1.0000 430
49
+ औ 1.0000 1.0000 1.0000 408
50
+ क 1.0000 1.0000 1.0000 433
51
+ क्ष 1.0000 1.0000 1.0000 480
52
+ ख 1.0000 1.0000 1.0000 456
53
+ ग 1.0000 1.0000 1.0000 444
54
+ घ 1.0000 1.0000 1.0000 480
55
+ च 1.0000 1.0000 1.0000 463
56
+ छ 1.0000 1.0000 1.0000 468
57
+ ज 1.0000 1.0000 1.0000 480
58
+ ज्ञ 1.0000 1.0000 1.0000 480
59
+ झ 1.0000 1.0000 1.0000 480
60
+ ट 1.0000 1.0000 1.0000 480
61
+ ठ 1.0000 1.0000 1.0000 480
62
+ ड 1.0000 1.0000 1.0000 480
63
+ ढ 1.0000 1.0000 1.0000 480
64
+ ण 1.0000 1.0000 1.0000 480
65
+ त 1.0000 1.0000 1.0000 480
66
+ थ 1.0000 1.0000 1.0000 480
67
+ द 1.0000 0.9875 0.9937 480
68
+ ध 1.0000 1.0000 1.0000 480
69
+ न 1.0000 1.0000 1.0000 480
70
+ प 1.0000 1.0000 1.0000 480
71
+ फ 1.0000 1.0000 1.0000 480
72
+ ब 1.0000 1.0000 1.0000 480
73
+ भ 1.0000 1.0000 1.0000 480
74
+ म 1.0000 1.0000 1.0000 480
75
+ य 1.0000 1.0000 1.0000 480
76
+ र 1.0000 1.0000 1.0000 484
77
+ ल 1.0000 1.0000 1.0000 480
78
+ ळ 1.0000 1.0000 1.0000 480
79
+ व 1.0000 1.0000 1.0000 480
80
+ श 1.0000 1.0000 1.0000 480
81
+ स 1.0000 1.0000 1.0000 480
82
+ ह 1.0000 1.0000 1.0000 480
83
+
84
+ accuracy 0.9997 20040
85
+ macro avg 0.9997 0.9997 0.9997 20040
86
+ weighted avg 0.9997 0.9997 0.9997 20040
87
+ ```
88
+ ![result.png]()
89
+ ## Training procedure
90
+
91
+ ### Training hyperparameters
92
+
93
+ The following hyperparameters were used during training:
94
+ - learning_rate: 2e-06
95
+ - train_batch_size: 32
96
+ - eval_batch_size: 8
97
+ - seed: 42
98
+ - optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
99
+ - lr_scheduler_type: linear
100
+ - lr_scheduler_warmup_steps: 50
101
+ - num_epochs: 6
102
+
103
+ ### Training results
104
+
105
+ | Training Loss | Epoch | Step | Validation Loss | Model Preparation Time | Accuracy |
106
+ |:-------------:|:-----:|:----:|:---------------:|:----------------------:|:--------:|
107
+ | 1.4439 | 1.0 | 940 | 0.0090 | 0.0057 | 0.9980 |
108
+ | 0.0052 | 2.0 | 1880 | 0.0035 | 0.0057 | 0.9993 |
109
+ | 0.0031 | 3.0 | 2820 | 0.0016 | 0.0057 | 0.9997 |
110
+ | 0.001 | 4.0 | 3760 | 0.0010 | 0.0057 | 0.9997 |
111
+ | 0.0007 | 5.0 | 4700 | 0.0013 | 0.0057 | 0.9997 |
112
+ | 0.0005 | 6.0 | 5640 | 0.0006 | 0.0057 | 0.9997 |
113
+
114
+
115
+ ### Framework versions
116
+
117
+ - Transformers 4.52.0.dev0
118
+ - Pytorch 2.6.0+cu124
119
+ - Datasets 3.5.1
120
+ - Tokenizers 0.21.1