https://www.kaggle.com/competitions/sai-vessel-segmentation/overview
Diabetic Retinopathy (DR) is a consequence of Type-II diabetes, resulting in harm to the blood vessels in the retina and contributing to visual impairment. The anticipated focus revolves around the likelihood of fluctuations in estimating retinal veins, with precision detecting vessel development within the retina.
In order to prevent the database from being deleted, I first backed it up to my personal cloud space. Download Link
- tensorflow-21.02-tf1-py3:latest
sudo apt-get update
sudo apt-get install ffmpeg libsm6 libxext6 -y
pip install imgaug
pip install -U scikit-learn scipy matplotlib
pip install pandas
python showimg.py
-
This code can be trained on Google Colab.
-
Augmentation lets the model more robust.
-
With the early stop callback mechanism.
-
My trained Weight: Download link
- This code directly uses Google Colab for inference images.
- Inference with different confidence thresholds:
Finished sixth in the competition with a dice value of 0.78, out of a first-place score of 0.80.