How sivapanyam/image_captioning_model works
This repository is currently a Flickr8k data-preparation scaffold for an image-captioning project, not a complete caption-generation system. Its implemented workflow configures dataset paths, creates and reads image splits, checks dataset consistency, and reports caption statistics. The README also describes the intended ResNet50-to-LSTM training, evaluation, and inference capabilities, but their implementation directories are absent from the tree. Those documented stages are shown as planned; their internal wiring is not evidenced here.
Main entry points: config.py Dataset config · check_dataset.py Dataset verifier · create_splits.py Split creator · Flickr8k Image split files
Components 16
Configuration
- Dataset config
configs/config.py
Dataset preparation
- Caption annotations
- Split creator
src/preprocessing/create_splits.py - Image split files
dataset/Flickr8k - Split reader
src/preprocessing/read_splits.py - Dataset verifier
src/preprocessing/check_dataset.py - Path verifier
src/utils/file_utils.py - Caption inspector
src/preprocessing/read_captions.py - Caption statistics
src/preprocessing/dataset_stats.py
Captioning system (planned)
- ResNet50 encoder
- LSTM decoder
- Model training
- Caption evaluation
- Caption inference
- Generated caption
Other components
- Dataset maintainer
Connections 20
- Dataset maintainer → Split creator: creates splits
src/preprocessing/create_splits.py - Split creator → Dataset config: reads paths
src/preprocessing/create_splits.py - Split creator → Caption annotations: reads annotations
src/preprocessing/create_splits.py - Split creator → Image split files: writes splits
src/preprocessing/create_splits.py - Split reader → Dataset config: reads paths
src/preprocessing/read_splits.py - Split reader → Image split files: reads splits
src/preprocessing/read_splits.py - Dataset verifier → Dataset config: reads paths
src/preprocessing/check_dataset.py - Dataset verifier → Path verifier: verifies paths
src/preprocessing/check_dataset.py - Dataset verifier → Caption annotations: counts annotations
src/preprocessing/check_dataset.py - Dataset verifier → Image split files: checks counts
src/preprocessing/check_dataset.py - Caption inspector → Dataset config: reads path
src/preprocessing/read_captions.py - Caption inspector → Caption annotations: prints samples
src/preprocessing/read_captions.py - Caption statistics → Dataset config: reads paths
src/preprocessing/dataset_stats.py - Caption statistics → Caption annotations: counts captions
src/preprocessing/dataset_stats.py - Model training → ResNet50 encoder: uses encoder
README.md - Model training → LSTM decoder: uses decoder
README.md - Caption evaluation → LSTM decoder: evaluates captions
README.md - Caption inference → ResNet50 encoder: extracts features
README.md - Caption inference → LSTM decoder: generates text
README.md - Caption inference → Generated caption: produces
README.md