main.py

import os
import io
from flask import Flask, request, jsonify
import base64
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

os.environ["TF_ENABLE_ONEDNN_OPTS"] = "0"

app = Flask(__name__)

# MAX_LENGTH = 34
MAX_LENGTH = 38

def extract_features(image_data, model):
    try:
        image = Image.open(io.BytesIO(image_data))
    except Exception as e:
        print("ERROR: Can't open image! Ensure that image data is correct and in the expected format")
        print(str(e))
        return None

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

    # for 4 channels images, we need to convert them 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']
        # return jsonify({'caption': base64_image_data}), 200
        # Replace spaces with "+" characters to handle cases where "+" characters are missing
        # base64_image_data = base64_image_data.replace(" ", "+")
        
        # Decode the Base64 string into binary image data
        image_data = base64.b64decode(base64_image_data)

        tokenizer = load(open("tokenizer.p","rb"))
        # model = load_model('model_9.h5')
        model = load_model('models/model_9.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
        
        caption = generate_desc(model, tokenizer, photo, MAX_LENGTH)

        # Return the generated caption
        return jsonify({'caption': caption}), 200
    except Exception as e:
        return jsonify({'error': str(e)}), 500

if __name__ == '__main__':
    app.run(host='0.0.0.0')