Generate_Captions.py

import matplotlib.pyplot as plt
import seaborn as sns
from PIL import Image
%matplotlib inline

# method for generating captions
def generate_captions(model, image, tokenizer.word_index, max_caption_length, tokenizer.index_word):
    
    # input is <start>
    input_text = '<start>'
    
    # keep generating words till we have encountered <end>
    for i in range(max_caption_length):
        seq = [tokenizer.word_index[w] for w in in_text.split() if w in list(tokenizer.word_index.keys())]
        seq = pad_sequences([sequence], maxlen=max_caption_length)
        prediction = model.predict([photo,sequence], verbose=0)
        prediction = np.argmax(prediction)
        word = tokenizer.index_word[prediction]
        input_text += ' ' + word
        if word == '<end>':
            break
    
    # remove <start> and <end> from output and return string
    output = in_text.split()
    output = output[1:-1]
    output = ' '.join(output)
    return output

# traverse through testing images to generate captions
count = 0
for key, value in test_image_features.items():
    test_image = test_image_features[key]
    test_image = np.expand_dims(test_image, axis=0)
    final_caption = generate_captions(predictive_model, test_image, tokenizer.word_index, max_caption_len, tokenizer.index_word)
    
    plt.figure(figsize=(7,7))
    image = Image.open(image_path + "//" + key + ".jpg")
    plt.imshow(image)
    plt.title(final_caption)
    
    count = count + 1
    if count == 3:
        break