Image Captioning - Monorepo

Full-stack app for captioning images with a small TorchScript model, guest sessions, and user accounts.

• Frontend: Next.js + TypeScript + shadcn/ui
• Backend: FastAPI + SQLModel (SQLite by default) + PyTorch (TorchScript)
• Auth: Guest token (cookie) + Bearer token; guest history kept 24h and merged on sign-in.

Repo Structure

• backend/ - FastAPI API, auth, inference, database
• frontend/ - Next.js UI
• model/ - notebooks, artifacts, training data
• scripts/ - Postgres bootstrap for container runtime
• Dockerfile, nginx.conf, supervisord.conf - container build and runtime

Live Demo

• UI: https://imagecaption.sahilkhadtare.com
• API: https://imagecaption.sahilkhadtare.com/api

Quick Start (local)

1. Backend

• Follow backend/README.md.

2. Frontend

• Follow frontend/README.md.

Model Artifacts

The backend needs TorchScript artifacts to run inference. Use the files in model/artifacts and copy them into backend/artifacts, or set MODEL_TS, VOCAB_JSON, and CONFIG_JSON to point to them.

See model/README.md for details.

Docker (optional)

The Docker image builds the frontend as static assets and runs the backend behind nginx. If you use the container, make sure backend/artifacts is present in the build context or set artifact environment variables at runtime.

docker build -t image-caption .
docker run -p 8080:8080 image-caption

• UI: http://localhost:8080
• API: http://localhost:8080/api