triopood commited on
Commit
4dcc6ef
·
verified ·
1 Parent(s): cd007be

Create app.py

Browse files
Files changed (1) hide show
  1. app.py +51 -0
app.py ADDED
@@ -0,0 +1,51 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ from flask import Flask, request, jsonify
2
+ from PIL import Image
3
+ import torch
4
+ from torchvision.transforms import functional as F
5
+ from transformers import TrOCRProcessor, VisionEncoderDecoderModel
6
+
7
+ app = Flask(__name__)
8
+
9
+ # Load the trained model and processor
10
+ processor = TrOCRProcessor.from_pretrained("microsoft/trocr-small-printed")
11
+ model = VisionEncoderDecoderModel.from_pretrained("trocrnew.pth")
12
+
13
+ # Set the model in evaluation mode
14
+ model.eval()
15
+
16
+ def ocr(image):
17
+ # Preprocess the image
18
+ image = F.to_tensor(image).unsqueeze(0)
19
+
20
+ # Perform OCR
21
+ with torch.no_grad():
22
+ generated_ids = model.generate(image)
23
+
24
+ # Decode the generated text
25
+ generated_text = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
26
+
27
+ return generated_text
28
+
29
+ @app.route('/ocr', methods=['POST'])
30
+ def ocr_endpoint():
31
+ # Check if the POST request contains a file
32
+ if 'file' not in request.files:
33
+ return jsonify({'error': 'No file provided'}), 400
34
+
35
+ file = request.files['file']
36
+
37
+ # Check if the file has an allowed extension
38
+ allowed_extensions = {'png', 'jpg', 'jpeg', 'gif'}
39
+ if '.' not in file.filename or file.filename.split('.')[-1].lower() not in allowed_extensions:
40
+ return jsonify({'error': 'Invalid file type'}), 400
41
+
42
+ # Read the image and perform OCR
43
+ try:
44
+ image = Image.open(file).convert('RGB')
45
+ text = ocr(image)
46
+ return jsonify({'text': text}), 200
47
+ except Exception as e:
48
+ return jsonify({'error': str(e)}), 500
49
+
50
+ if __name__ == '__main__':
51
+ app.run(debug=True)