main.py

import io
from flask import Flask, request, jsonify
import base64
import pytesseract
import numpy as np
from pickle import load
from PIL import Image
from keras.applications.xception import Xception # to get pre-trained model Xception
from keras.models import load_model
from keras.preprocessing.sequence import pad_sequences

app = Flask(__name__)

MAX_LENGTH = 34

def format_tesseract_output(output_text):
    formatted_text = ""
    lines = output_text.strip().split("\n")
    for line in lines:
        line = line.strip()
        if line:
            formatted_text += line + "\n"
    return formatted_text


def extract_features(image_data, model):
    try:
        image = Image.open(io.BytesIO(image_data))
    except Exception as e:
        return None

    image = image.resize((299,299))
    image = np.array(image)

    # convert 4 channels image into 3 channels
    if image.shape[2] == 4:
        image = image[..., :3]

    image = np.expand_dims(image, axis=0)
    image = image/127.5
    image = image - 1.0
    feature = model.predict(image)

    return feature


def word_for_id(integer, tokenizer):
    for word, index in tokenizer.word_index.items():
        if index == integer:
            return word
    return None


def generate_desc(model, tokenizer, photo, max_length):
    in_text = 'start'
    for i in range(max_length):
        sequence = tokenizer.texts_to_sequences([in_text])[0]
        sequence = pad_sequences([sequence], maxlen=max_length)
        pred = model.predict([photo,sequence], verbose=0)
        pred = np.argmax(pred)
        word = word_for_id(pred, tokenizer)
        if word is None or word == 'end':
            break
        in_text += ' ' + word
    return in_text.replace('start ', '')


# API endpoint to receive image and generate caption
@app.route('/api', methods=['POST'])
def generate_caption():
    try:
        base64_image_data = request.form['image']
        
        # Decode the Base64 string into binary image data
        image_data = base64.b64decode(base64_image_data)

        # Convert the image data to a PIL image object
        pil_image = Image.open(io.BytesIO(image_data))
        
        extracted_text = pytesseract.image_to_string(pil_image, lang="eng+chi_sim+msa")
        hasText = bool(extracted_text.strip())
        
        if hasText:
            result = format_tesseract_output(extracted_text)
        else:
            tokenizer = load(open("tokenizer.p","rb"))
            model = load_model('model.keras')

            xception_model = Xception(include_top=False, pooling="avg")
            photo = extract_features(image_data, xception_model)

            if photo is None:
                return jsonify({'error': 'Failed to extract features from the image'}), 400
        
            result = generate_desc(model, tokenizer, photo, MAX_LENGTH)

        return jsonify({'hasText': hasText, 'result': result}), 200
    except Exception as e:
        return jsonify({'error': str(e)}), 500

if __name__ == '__main__':
    app.run()