KO
|
EN
gitlite — search
Search
#javascript
#python
#hacktoberfest
#react
#ai
#typescript
#llm
#go
#golang
#android
#machine-learning
#rust
#deep-learning
#linux
GCN-CIS
★ 14
Open GitHub ↗
No description available.
Download README (.md)
Explore Similar Repositories
prostNfound
:
Prostate cancer detection from ultrasound using foundation models and domain knowledge
wave-function-collapse-ruby
:
Tiled Wave Function Collapse model implemented in Ruby and rendered using Gosu
GKR-MSM
:
No description available.
Multiple-Object-Tracking-system
:
系统支持：多目标实时跟踪，上传视频跟踪、视频管理、目标计数、跟踪截图报告、跟踪结果下载、远程域名访问。 多目标跟踪算法检测部分是：YOLOv8。数据关联部分：重新设计的一个算法。系统框架：Django。系统做了内网穿透支持远程域名访问。视频实时推流：重新写了一个视频推流方法，延迟能做到300ms左右，基本能满足实时跟踪，偶尔会出现掉帧现象。
PKU2023fall-VCL-lab
:
This is the code for the PKU2023fall VCL lab.
// repository documentation
Was this content helpful?
★ 0
(0 ratings)
Select Rating:
★
★
★
★
★
Submit Feedback
Recent Feedback
×
Download README
Do you want to download the
README.md
file for
GCN-CIS
?
Download (.md)
# Geometric Constraint Network for Clinically Interpretable Optic Cup and Disc Segmentation (GCN-CIS) ## ✈️ Highlights 🔥 A novel segmentation network named GCN-CIS, which is enhanced with a Prior Boundary Attention Module. 🔥 A Geometrically Constrained Loss function tailored specifically to the unique morphological characteristics of the optic cup and disc. 🔥 GCN-CIS achieves outstanding performance on our self-constructed Glaucoma Screening Dataset and three other public datasets. 🔥 A Website based on GCN-CIS can assist doctors in diagnosis [RetinalSeg Expert](https://glaucoma-assistant.com/cdr/). ## 👉 Framework <img src="https://github.com/QinRui-k/GCN-CIS/assets/139854014/4e46819a-3274-4548-aca7-2f28ac8159a4"> Architecture of GCN-CIS. The network incorporates a multi-scale input strategy to capture details across various resolutions. It utilizes a series of encoders for extracting features at multiple levels of abstraction. Subsequently, the deep feature maps undergo refinement through the Prior Boundary Attention Module (PBAM), which significantly enhances edge delineation. These refined features are then integrated into the decoder, thereby improving the representation of the optic boundaries. Notably, the final segmentation map is crucial for calculating the signed distance and assessing the ISNT metrics of the image, with an integrated loss function that sharpens the segmentation accuracy. ## 👉 Evaluation <img src="https://github.com/QinRui-k/GCN-CIS/assets/139854014/764d7de2-c741-4f5a-9f45-52cb7446ab71"> <p align="center"> Visulization Results </p> <img src="https://github.com/QinRui-k/GCN-CIS/assets/139854014/ac25f5e3-78ef-4d43-a0d5-2522048fb9ee"> <p align="center"> Quantitative Results </p> ## 👉 Train ``` cd ./GCN-CIS train.py ``` ## 👉 Test ``` cd ./GCN-CIS test.py ``` ## 🎥 Demo https://github.com/QinRui-k/GCN-CIS/assets/139854014/eb36b117-ba0d-4068-a0bf-5cae07e52ed8 ## 🤝 Acknowledgement * We thank all medical workers and dataset owners for making public datasets available to the community. * We thank public dataset owner [Drishti-GS](https://www.kaggle.com/datasets/lokeshsaipureddi/drishtigs-retina-dataset-for-onh-segmentation), [REFUGE](https://refuge.grand-challenge.org) and [ORIGA](https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=5626137).
