A longer video with audio can be accessed here.
The model is too large to store on github (165 mb), download it from this gdrive link and move it into the base directory instead. I am working on a way to deal with the issue.
The KKP Blood Screening Project is a web application designed for blood analysis using YOLOv5 object detection model deployed with Flask. This application allows users to upload an image and receive a labeled result indicating various blood cell types. The model is trained to detect 11 classes of blood cells, including:
- Lymphoblast: 752 instances
- Neutrophil: 263 instances
- Promyelocytes: 137 instances
- Dysplastic_Cell: 6 instances
- Lymphocytes: 4 instances
- Myelocytes: 4 instances
- Band_Neutrophils: 2 instances
- Nucleated_RBC: 2 instances
- Giant_Platelet: 1 instance
- Monocytes: 1 instance
For more detailed information about these cell types, refer to the blood_detect folder.
To run the project locally for development purposes, follow these steps:
- Create a virtual environment:
python3 -m venv venv
- Activate the virtual environment:
source venv/bin/activate - Install the required dependencies:
(venv) $ pip install -r requirements.txt
- Run the Flask web application:
(venv) $ python webapp.py
Additionally, an example Python script for performing inference using requests is provided in tests/test_request.py. Please note that the find_model() function within this script expects a file ending with .pt for inference, and the code will work only if there is one model file ending with .pt placed inside the folder. Adjust the script accordingly if multiple model files are present.
This project is licensed under the MIT License.
- This project utilizes the YOLOv5 model developed by Ultralytics.