HACDR-Net: Heterogeneous-Aware Convolutional Network for Diabetic Retinopathy Multi-Lesion Segmentation
Authors: QiHao Xu, Xiaoling Luo, Chao Huang, Chengliang Liu, Jie Wen, Jialei Wang, Yong Xu
AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We conduct the experiments on the public datasets IDRi D and DDR, and the experimental results show that the proposed method achieves better performance than other state-of-the-art methods. |
| Researcher Affiliation | Academia | 1College of Computer Science and Software Engineering, Shenzhen University, Shenzhen, China 2Shenzhen Key Laboratory of Visual Object Detection and Recognition, Harbin Institute of Technology, Shenzhen, China 3School of Cyber Science and Technology, Shenzhen Campus of Sun Yat-sen University, Shenzhen 518107, China |
| Pseudocode | No | The paper describes the methods using text and diagrams but does not contain structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | The code is opensourced on github.com/xqh180110910537/HACDR-Net. |
| Open Datasets | Yes | Two publicly available DR multi-lesion segmentation datasets are adopted, i.e., the Indian Diabetic Retinopathy Image Dataset (IDRi D), A General-purpose High-quality Dataset for Diabetic Retinopathy Classification, Lesion Segmentation and Lesion Detection (DDR). DDR: The DDR (Li et al. 2019) dataset contains 757 images of fundus lesions with pixel-level annotations, including 383 images for training, 149 images for validation, and 225 images for testing. IDRi D: The IDRi D (Porwal et al. 2018) dataset only contains 81 images of fundus lesions with pixel-level annotations, including 54 images for training and 27 images for testing. |
| Dataset Splits | Yes | DDR: The DDR (Li et al. 2019) dataset contains 757 images of fundus lesions with pixel-level annotations, including 383 images for training, 149 images for validation, and 225 images for testing. IDRi D: The IDRi D (Porwal et al. 2018) dataset only contains 81 images of fundus lesions with pixel-level annotations, including 54 images for training and 27 images for testing. |
| Hardware Specification | Yes | All models are trained on a node with 2 RTX 3090 GPUs. |
| Software Dependencies | No | The paper states, "Our implementation is based on mmsegmentation (Contributors 2020) libraries," but does not provide specific version numbers for mmsegmentation or other software dependencies. |
| Experiment Setup | Yes | The batch size is set to 1∼4 according to different resolutions for these two datasets. Adam W (Loshchilov and Hutter 2017) is applied to train our models. We set the initial learning rate as 0.00006 and employ the poly-learning rate decay policy. |