CNN-LSTM Image Captioning

An image-captioning system combining convolutional visual representations with LSTM-based language generation.

Architecture

• VGG16: Visual and spatial feature extraction from the input image.
• LSTM: Sequence modelling and caption generation based on extracted features.

Problem & Approach

Generating natural language descriptions of images requires translating spatial features into sequential tokens. By coupling a pre-trained CNN (VGG16) with a Recurrent Neural Network (LSTM), the system bridges computer vision and natural language processing.

Tech Stack

• TensorFlow
• Keras
• VGG16
• LSTM
• Computer Vision
• Deep Learning