Weakly Supervised Semantic Segmentation Using Superpixel Pooling Network

Authors: Suha Kwak, Seunghoon Hong, Bohyung Han

AAAI 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental The proposed algorithm achieves outstanding performance in weakly supervised semantic segmentation task compared to existing techniques on the challenging PASCAL VOC 2012 segmentation benchmark.
Researcher Affiliation Academia 1Department of Information and Communication Engineering, DGIST, Korea 2Department of Computer Science and Engineering, POSTECH, Korea
Pseudocode No The paper includes an architectural diagram (Figure 1) and mathematical derivations, but no explicitly labeled pseudocode or algorithm blocks.
Open Source Code No The paper does not provide any explicit statement about releasing source code or a link to a code repository for the methodology described.
Open Datasets Yes Both of SPN and Decoupled Net are trained on PASCAL VOC 2012 dataset (Everingham et al. 2010). Besides the provided image sets for the semantic segmentation task, we employ additional images used in (Hariharan et al. 2011) to enlarge training set.
Dataset Splits Yes In total, 10,582 images are used to train the networks, and the validation set of 1,449 images is kept for evaluating our approach.
Hardware Specification Yes The training procedure takes about 3 hours on a single Nvidia TITAN X GPU with 12Gb RAM in our experiment.
Software Dependencies Yes SPN is implemented in Torch7 (Collobert, Kavukcuoglu, and Farabet 2011).
Experiment Setup Yes The network parameters are optimized by Adam (Ba and Kingma 2015) with an initial learning rate of 0.001. The optimization converges after approximately 9K iterations with mini-batches of 12 images.