https://github.com/rishic3/strepdetection
An interpretable deep learning approach to detect strep throat directly from cell phone videos.
https://github.com/rishic3/strepdetection
Last synced: about 2 months ago
JSON representation
An interpretable deep learning approach to detect strep throat directly from cell phone videos.
- Host: GitHub
- URL: https://github.com/rishic3/strepdetection
- Owner: rishic3
- Created: 2023-12-01T18:04:10.000Z (over 2 years ago)
- Default Branch: main
- Last Pushed: 2025-01-16T19:19:37.000Z (over 1 year ago)
- Last Synced: 2025-04-01T21:18:08.240Z (about 1 year ago)
- Language: Python
- Homepage:
- Size: 70.3 MB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# StrepDetection
Detection of strep throat directly from cell phone videos.
Employing intermediate symptom classification combined with rule-based decisions for interpretable results.
Implementing strategies (hard-negative mining, contrastive learning) to combat limited and imbalanced data.
## Parsing data from CVAT:
1. Download data from CVAT
* `Actions > Export Dataset > Export Format: CVAT for video 1.1`.
* This will download a folder containing an xml file with the dataset annotations.
3. Parse annotations via `parse_xml.py`
* Set the xml file path and run `parse_xml.py`.
* This will produce a .csv file with the video, frame, and relevant labels.
4. Merge CVAT data with .xlsx data
* Follow the steps in `data_process.ipynb`.
* This will merge the annotations from the `.xlsx` training review with the CVAT labels, checking for any overlap.
## Model Checkpoints:
[OneDrive folder](https://livejohnshopkins-my.sharepoint.com/:f:/g/personal/rchand18_jh_edu/Eqpi0aQnp_ZNmqp5sNe990EBUEEEuu3CyJAAGzhS831qXQ?e=kVGbqF) containing model checkpoints.
Authored by Rishi Chandra, rchand18@jhu.edu, as part of the [ARCADE Lab](https://arcade.cs.jhu.edu/) at Johns Hopkins University.