fork download
copy 
  1. import numpy as np
  2. import tensorflow as tf
  3. from tensorflow.keras.applications import InceptionV3
  4. from tensorflow.keras.models import Model, load_model
  5. from tensorflow.keras.preprocessing import image
  6. from tensorflow.keras.applications.inception_v3 import preprocess_input
  7. from tensorflow.keras.preprocessing.sequence import pad_sequences
  8. from tensorflow.keras.layers import Embedding, LSTM, Dense
  9.  
  10. # Load pre-trained InceptionV3 model for feature extraction
  11. def load_cnn_model():
  12.     base_model = InceptionV3(weights='imagenet')
  13.     cnn_model = Model(base_model.input, base_model.layers[-2].output)
  14.     return cnn_model
  15.  
  16. # Extract image features using CNN
  17. def extract_image_features(cnn_model, img_path):
  18.     img = image.load_img(img_path, target_size=(299, 299))
  19.     img_array = image.img_to_array(img)
  20.     img_array = np.expand_dims(img_array, axis=0)
  21.     img_array = preprocess_input(img_array)
  22.     features = cnn_model.predict(img_array)
  23.     return features
  24.  
  25. # Load tokenizer and pre-trained captioning model
  26. def load_captioning_model(tokenizer_path, captioning_model_path):
  27.     tokenizer = tf.keras.models.load_model(tokenizer_path)
  28.     captioning_model = load_model(captioning_model_path)
  29.     return tokenizer, captioning_model
  30.  
  31. # Generate caption from image features
  32. def generate_caption(features, tokenizer, captioning_model, max_length):
  33.     caption = 'startseq'
  34.     for _ in range(max_length):
  35.         sequence = tokenizer.texts_to_sequences([caption])[0]
  36.         sequence = pad_sequences([sequence], maxlen=max_length)
  37.         pred_word_index = np.argmax(captioning_model.predict([features, sequence]), axis=-1)
  38.         word = tokenizer.index_word.get(pred_word_index[0], '')
  39.         if word == 'endseq' or not word:
  40.             break
  41.         caption += ' ' + word
  42.     return caption.replace('startseq', '').strip()
  43.  
  44. # Main function
  45. def main():
  46.     cnn_model = load_cnn_model()
  47.     img_path = 'sample_image.jpg'  # Provide the path to your sample image
  48.     tokenizer_path = 'tokenizer.pkl'  # Path to the saved tokenizer
  49.     captioning_model_path = 'captioning_model.h5'  # Path to the saved captioning model
  50.     max_length = 20  # Max caption length
  51.  
  52.     # Extract image features
  53.     features = extract_image_features(cnn_model, img_path)
  54.  
  55.     # Load tokenizer and captioning model
  56.     tokenizer, captioning_model = load_captioning_model(tokenizer_path, captioning_model_path)
  57.  
  58.     # Generate caption
  59.     caption = generate_caption(features, tokenizer, captioning_model, max_length)
  60.     print("Generated Caption:", caption)
  61.  
  62. if __name__ == '__main__':
  63.     main()
  64.  
Time limit exceeded #stdin #stdout #stderr 5s 269668KB
comments (?)
stdin
copy 
Standard input is empty
stdout
copy 
Standard output is empty
stderr
copy 
WARNING:tensorflow:From /usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/resource_variable_ops.py:435: colocate_with (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version.
Instructions for updating:
Colocations handled automatically by placer.
https://ideone.com/qUT6zv
language:
Python (cpython 2.7.16)
created:
13 hours ago
visibility:
public

Share or Embed source code

Discover > Sphere Engine API

The brand new service which powers Ideone!

Discover > IDE Widget

Widget for compiling and running the source code in a web browser!