This application is designed for doctors to review and validate AI-generated bounding box annotations for MRI scans.
git clone https://github.com/ViVi-Team/data-app.git
cd data-appSince this project uses Git LFS for large image files, you must initialize it:
git lfs install
git lfs pullAlternatively, you can download the results data manually using gdown:
pip install gdown
gdown --id 1fj7xZWUp9YyMwVSCcyEbbQHtCHOa08hC -O results.zip
unzip results.zippython3 -m venv .venv
source .venv/bin/activate # On Windows use `.venv\Scripts\activate`
pip install -r requirements.txtRun the following command in your terminal:
streamlit run review_app.py- Enter your Name or ID in the sidebar on the left.
- Click the Login button.
Use the dropdown menu in the sidebar to select a patient from the list.
- Use the ⬅ Prev and Next ➡ buttons at the top to move through the MRI slides.
- You can also use your Left and Right arrow keys on your keyboard.
Each slide displays a 3-panel view:
- Left (Original): The raw MRI slide for reference.
- Center (Gold Source): The original "Gold Standard" slide that the AI used for propagation.
- Right (Perturbed): The AI-generated annotation on the current slide.
- Approve [A]: If the AI-generated box on the right correctly identifies the region.
- Reject [R]: If the box is incorrect or misplaced.
- Reset [C]: Clear your decision for the current slide.
Note
Gold Standard slides (the original manual annotations) will have the Approve/Reject buttons disabled, as they are already verified.
The app saves your decisions locally in your browser session as you work.
When you have finished your review:
- Click 📊 Export Final Results (CSV) in the sidebar to download your decisions.
- If you need to pause and resume later on another machine, click 📥 Export Session (JSON) to save your progress, and use the 📤 Import Session box to restore it later.
| Action | Key |
|---|---|
| Approve | A |
| Reject | R |
| Reset | C |
| Previous Slide | Left Arrow |
| Next Slide | Right Arrow |