English | 简体中文
Please star this project in the upper right corner and cite this paper blow if this project helps you.
This repository is the source codes for the paper "Optimizing Gastrointestinal Polyp Recognition through Low-Redundancy Dataset".
GapoNet (Gastrointestinal Polyp Detection Network) is proposed to detect gastrointestinal polyps for endoscope images.
Graphical abstract.
-
A pretrained high-dimensional feature-driven cosine similarity deduplication technique was proposed to construct GapoSet, an AI dataset for gastrointestinal polyp recognition, characterized by low similarity and the great diversity.
-
Detection was conducted using the improved YOLO11 network (with CA and MHSA attention mechanisms), and state-of-the-art performance was achieved.
Get GapoNet code and configure the environment, please check out docs/INSTALL.md
Download datasets and trained weights, please check out docs/DATASETS.md
python val.py
The main optional arguments:
--model "xx.pt"
--data "xx.yaml"
--device "0, 1" # cpu or gpu id
--imgsz 640
--batch 32
--base_save_dir "xx"
Once you get the LymoNet code, configure the environment and download the dataset, just type:
python train.py
The training results and weights will be saved in runs/detect/directory.
The main optional arguments:
--model "x.yaml"
--data "xx.yaml"
--device "0, 1" # cpu or gpu id
--imgsz 640
--batch 32
--epochs 300
GapoNet is authored by Menglu Tan, Zhengde Zhang, Ao Wang, Zijin Zeng, and Lin Feng.
Currently, it is maintained by Menglu Tan ([email protected]).
This work was supported by the Beijing Municipal Fund for Distinguished Young Scholars (Grand No. JQ22022), National Key R&D Program of China (Grant No. 2022YFF1502000).
We are very grateful to the ultralytics project for the benchmark detection algorithm.
GapoNet is freely available for free non-commercial use, and may be redistributed under these conditions. For commercial queries, please drop an e-mail at [email protected]. We will send the detail agreement to you.