Computer vision delivery path
ViT Pet Classification Pipeline
This project packages a compact ML delivery path around ViT fine-tuning and inference. It includes data handling, training/evaluation scripts, CLI prediction, FastAPI serving, and a Streamlit UI while keeping datasets and model weights out of version control.
Highlights
- Keeps heavy datasets and model artifacts outside version control.
- Supports CLI, API, and UI entry points for inference review.
- Documents repository posture and model assumptions.
Validation
5 pytest tests
GitHub Actions CI
REPORT.md
Sanitized UI screenshot
