According poisson image blending I've completely used it for biomedical image segmentation to increase dataset. Because Polyps segementation dataset does not enough have.
Wheares there are many tasks in medical using image to diagnose. This will help the doctors reduce time and pressure 💪
• I think this type is so cool!
You can see new image after copy paste really like a real picture 😄
git clone https://github.com/hungk64it1x/PIB-for-medical-image
pip install -r requirements.txt
or for Linux
sudo apt install -r requirements.txt
after that you should modify the image source (contain positive region) in file run.py from line 84 -> 89 because my implement that for Kvasir, Colon, Etis, Clinic and EndoScene dataset.
Finally run run.py to generate copy image:
python3 run.py --image_dir [directory of image] --mask_dir [directory of mask] --save_image_dir [path save images] --save_mask_dir [path save masks] --num_images [number of image to generate]
This research was implemented when I was in a research group funded by Vingroup Innovation Foundation conducted by PhD Sang D.V in HUST😄
- Poisson image editing
- HarDnet-MSEG
- Specially thank to github