Weakly-Supervised Mirror Detection via Scribble Annotations

Authors: Mingfeng Zha, Yunqiang Pei, Guoqing Wang, Tianyu Li, Yang Yang, Wenbin Qian, Heng Tao Shen

AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments on three mirror datasets show that our network outperforms relevant state-of-the-art methods on all evaluation metrics and achieves performance comparable to fully supervised approaches.
Researcher Affiliation Academia Mingfeng Zha1, Yunqiang Pei1, Guoqing Wang1*, Tianyu Li1, Yang Yang1, Wenbin Qian2, Heng Tao Shen1 1University of Electronic Science and Technology of China 2Jiangxi Agricultural University
Pseudocode No The paper describes the proposed modules and their operations using mathematical formulas and text, but it does not include any pseudocode or clearly labeled algorithm blocks.
Open Source Code Yes The dataset and codes are available at https://github.com/winter-flow/WSMD.
Open Datasets Yes We collect training images from MSD, PMD, and Mirror-RGBD datasets, totaling 10,158 images, and relabel them as the training set of S-Mirror dataset. Models are evaluated using the testing sets of the above three datasets.
Dataset Splits No The paper specifies training and testing sets, but does not explicitly mention a separate validation set or its split details.
Hardware Specification Yes We implement our network using Py Torch and conduct experiments on an A100 GPU.
Software Dependencies No The paper only mentions "Py Torch" without specifying its version or other software dependencies with their respective version numbers.
Experiment Setup Yes All images are resized to 352 x 352. During the training phase, the batch size is 16, the initial learning rate is 1e-4, the decay rate is 0.9, Adam is used as the optimizer, and the epoch is 150.