sujay-mann/Text_Recognition

1from transformers import TrOCRProcessor, VisionEncoderDecoderModel2from PIL import Image3import streamlit as st4from torchvision.transforms.functional import pil_to_tensor5 6processor = TrOCRProcessor.from_pretrained('microsoft/trocr-base-handwritten')7model = VisionEncoderDecoderModel.from_pretrained('microsoft/trocr-base-handwritten')8 9def predict(image):10    print(image)11    if image.mode != 'RGB':12        image = image.convert('RGB')13    pixel_values = processor(images=image, return_tensors="pt").pixel_values14    generated_ids = model.generate(pixel_values)15    generated_text = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]16    return generated_text17 18def main():19    img_file = st.file_uploader(label='Image', type=['png', 'jpg', 'jpeg'])20    if img_file is not None:21        st.image(img_file, caption=img_file.name)22        image = Image.open(img_file)23        if st.button('Predict'):24            result = predict(image)25            st.write(f'Predicted Text: {result}')26        27if __name__ == '__main__':28    main()