W2P: Switching from Weak Supervision to Partial Supervision for Semantic Segmentation

Authors: Fangyuan Zhang, Tianxiang Pan, Jun-Hai Yong, Bin Wang

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

Reproducibility Variable Result LLM Response
Research Type Experimental Evaluations on the PASCAL VOC 2012 and MS COCO 2014 datasets confirm our method s impressive segmentation capabilities across various pseudo-labels.
Researcher Affiliation Academia 1School of Software, Tsinghua University, China 2Beijing National Research Center for Information Science and Technology (BNRist), China
Pseudocode No The paper describes algorithms and methods in prose and with equations, but it does not include any explicitly labeled "Pseudocode" or "Algorithm" blocks.
Open Source Code No The paper does not contain any statement about releasing source code or a link to a code repository.
Open Datasets Yes Our experiments use the benchmarks PASCAL VOC 2012 and MS COCO 2014 for WSSS.
Dataset Splits Yes PASCAL VOC 2012 has 10,582 training images, 1,449 validation images, and 1,456 testing images across 21 categories. MS COCO 2014 has 81 categories with 82,783 training images and 40,504 validation images.
Hardware Specification No The paper mentions using a "Res Net101 backbone" but does not specify any particular hardware (e.g., GPU models, CPU types, memory) used for running the experiments.
Software Dependencies No The paper mentions software like "Deeplab V3+" but does not provide specific version numbers for any software dependencies, libraries, or frameworks used in the experiments.
Experiment Setup Yes The W2P framework s hyperparameters need minimal tuning, with threshold values τ1 and τ2 set to 0.9 and λ set to 0.99. The BPND stage is trained for 8 epochs on VOC and 4 epochs on COCO, while the PSL stage takes 72 epochs on VOC and 36 epochs on COCO. A batch size of 16 is used for all experiments.