This project implements an image captioning system using a CNN as the image encoder and an LSTM-based RNN as the caption decoder. The model is trained on the COCO dataset to generate captions for a given image.
image-captioning/ │ ├── data/ # Store or link to your dataset │ ├── captions_train2017.json # COCO annotations (if downloading locally) │ └── images/ # Store images (if downloading locally) │ ├── models/ # Model files │ ├── encoder.py # CNN encoder module │ ├── decoder.py # RNN decoder module │ └── train.py # Training script │ ├── notebooks/ # Jupyter notebooks for exploration │ └── Image_Captioning.ipynb # Jupyter notebook demonstration │ ├── utils/ # Utility functions │ ├── vocab.py # Vocabulary processing │ ├── coco_dataset.py # COCO Dataset handling │ └── data_loader.py # DataLoader utilities │ ├── checkpoints/ # Store model checkpoints │ └── best_model.pth # Best model weights │ ├── requirements.txt # List of dependencies ├── README.md # Project description and setup instructions ├── train.py # Main training script └── eval.py # Script to evaluate model performance git clone https://github.com/yourusername/image-captioning.git cd image-captioning pip install -r requirements.txtTraining the Model To train the image captioning model, run the following command:
python3 train.py python3 eval.py