Update README.md
Browse files
README.md
CHANGED
@@ -85,7 +85,69 @@ Classification Report:
|
|
85 |
macro avg 0.9997 0.9997 0.9997 20040
|
86 |
weighted avg 0.9997 0.9997 0.9997 20040
|
87 |
```
|
88 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
89 |
## Training procedure
|
90 |
|
91 |
### Training hyperparameters
|
|
|
85 |
macro avg 0.9997 0.9997 0.9997 20040
|
86 |
weighted avg 0.9997 0.9997 0.9997 20040
|
87 |
```
|
88 |
+
---
|
89 |
+
|
90 |
+
## Install Dependencies
|
91 |
+
|
92 |
+
```bash
|
93 |
+
pip install -q transformers torch pillow gradio
|
94 |
+
```
|
95 |
+
|
96 |
+
---
|
97 |
+
|
98 |
+
## Inference Code
|
99 |
+
|
100 |
+
```python
|
101 |
+
import gradio as gr
|
102 |
+
from transformers import AutoImageProcessor, SiglipForImageClassification
|
103 |
+
from PIL import Image
|
104 |
+
import torch
|
105 |
+
|
106 |
+
# Load model and processor
|
107 |
+
model_name = "VinayHajare/siglip2-finetuned-marathi-sign-language"
|
108 |
+
model = SiglipForImageClassification.from_pretrained(model_name)
|
109 |
+
processor = AutoImageProcessor.from_pretrained(model_name)
|
110 |
+
|
111 |
+
# Marathi label mapping
|
112 |
+
id2label = {
|
113 |
+
"0": "अ", "1": "आ", "2": "इ", "3": "ई", "4": "उ", "5": "ऊ",
|
114 |
+
"6": "ए", "7": "ऐ", "8": "ओ", "9": "औ", "10": "क", "11": "क्ष",
|
115 |
+
"12": "ख", "13": "ग", "14": "घ", "15": "च", "16": "छ", "17": "ज",
|
116 |
+
"18": "ज्ञ", "19": "झ", "20": "ट", "21": "ठ", "22": "ड", "23": "ढ",
|
117 |
+
"24": "ण", "25": "त", "26": "थ", "27": "द", "28": "ध", "29": "न",
|
118 |
+
"30": "प", "31": "फ", "32": "ब", "33": "भ", "34": "म", "35": "य",
|
119 |
+
"36": "र", "37": "ल", "38": "ळ", "39": "व", "40": "श", "41": "स", "42": "ह"
|
120 |
+
}
|
121 |
+
|
122 |
+
def classify_marathi_sign(image):
|
123 |
+
image = Image.fromarray(image).convert("RGB")
|
124 |
+
inputs = processor(images=image, return_tensors="pt")
|
125 |
+
|
126 |
+
with torch.no_grad():
|
127 |
+
outputs = model(**inputs)
|
128 |
+
logits = outputs.logits
|
129 |
+
probs = torch.nn.functional.softmax(logits, dim=1).squeeze().tolist()
|
130 |
+
|
131 |
+
prediction = {
|
132 |
+
id2label[str(i)]: round(probs[i], 3) for i in range(len(probs))
|
133 |
+
}
|
134 |
+
|
135 |
+
return prediction
|
136 |
+
|
137 |
+
# Gradio Interface
|
138 |
+
iface = gr.Interface(
|
139 |
+
fn=classify_marathi_sign,
|
140 |
+
inputs=gr.Image(type="numpy"),
|
141 |
+
outputs=gr.Label(num_top_classes=5, label="Marathi Sign Classification"),
|
142 |
+
title="Marathi-Sign-Language-Detection",
|
143 |
+
description="Upload an image of a Marathi sign language hand gesture to identify the corresponding character."
|
144 |
+
)
|
145 |
+
|
146 |
+
if __name__ == "__main__":
|
147 |
+
iface.launch()
|
148 |
+
```
|
149 |
+
|
150 |
+
---
|
151 |
## Training procedure
|
152 |
|
153 |
### Training hyperparameters
|