MVCINN: Multi-View Diabetic Retinopathy Detection Using a Deep Cross-Interaction Neural Network
Authors: Xiaoling Luo, Chengliang Liu, Waikeung Wong, Jie Wen, Xiaopeng Jin, Yong Xu
AAAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments on the latest public multi-view MFIDDR dataset with 34,452 images demonstrate the superiority of our method, which performs favorably against state-of-the-art models. |
| Researcher Affiliation | Academia | 1Shenzhen Key Laboratory of Visual Object Detection and Recognition, Harbin Institute of Technology, Shenzhen, China 2School of Fashion and Textiles, The Hong Kong Polytechnic University, Kowloon, Hong Kong 3Laboratory for Artificial Intelligence in Design, Hong Kong 4College of Big Data and Internet, Shenzhen Technology University, Shenzhen, China |
| Pseudocode | Yes | Algorithm 1: The training process of MVCINN |
| Open Source Code | No | The paper does not explicitly state that the source code for their methodology is released or provide a link to it. The provided GitHub link (https://github.com/mfiddr/MFIDDR) is for the MFIDDR dataset, not the authors' code. |
| Open Datasets | Yes | We conducted experiments on the multi-field imaging dataset for DR detection (MFIDDR1), which is the only publicly available large-scale dataset of multi-view fundus images on DR so far. 1https://github.com/mfiddr/MFIDDR |
| Dataset Splits | No | The paper states, "Training and testing sets have been distributed on the MFIDDR, including 25,848 training images and 8,604 test images." It does not explicitly mention a separate validation set split or its size. |
| Hardware Specification | No | The paper does not provide specific hardware details such as GPU models, CPU specifications, or memory used for running the experiments. It only mentions general computing environments like "on Py Torch". |
| Software Dependencies | No | The paper mentions "Py Torch" as the framework and "Adam W optimizer" but does not specify any version numbers for these or other software components. |
| Experiment Setup | Yes | Implementation Details. The backbone of the convolutional branch with Conv-Block in our network is initialized by Res Net-50 (He et al. 2016) pre-trained on Image Net (Deng et al. 2009). The transformer branch with Trans Block is composed of a 12-layers transformer encoder with 9 heads, which is first pre-trained on the Image Net dataset. Furthermore, our model is achieved on Py Torch, and we use a random gradient coefficient with a base learning rate 1e 5 to improve our model. Our model chooses a learning rate of 1e 5, and use the Adam W optimizer (Kingma and Ba 2015) with the cosine annealing schedule (Loshchilov and Hutter 2017). Our model achieves much better performance when γ = 2. |