Computer Vision API
Several image models exposed through one service that other software could call.
Built a FastAPI application for image colorization, optical character recognition, and super resolution. It supported single and bulk uploads, background processing, a small web interface, and public deployment.
The point of the project was to make different image operations available through a predictable HTTP interface instead of asking a user to run separate notebooks.
The service
Each operation had its own processing path, while the API handled uploads, validation, and longer jobs in the background. The same service could return a processed image, recognized text, or an upscaled result.
What it taught me
A model is not an application by itself. Upload handling, response shape, background work, deployment, and failure cases decide whether another person can actually use it.
selected tools
- Python
- FastAPI
- OpenCV
- Tesseract
- ESRGAN