Balanced Meta-Softmax for Long-Tailed Visual Recognition

Authors: Jiawei Ren, Cunjun Yu, shunan sheng, Xiao Ma, Haiyu Zhao, Shuai Yi, hongsheng Li

NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In our experiments, we demonstrate that Balanced Meta-Softmax outperforms state-of-the-art long-tailed classification solutions on both visual recognition and instance segmentation tasks.
Researcher Affiliation Collaboration Jiawei Ren1, Cunjun Yu1, Shunan Sheng1,2, Xiao Ma1,3, Haiyu Zhao1*, Shuai Yi1, Hongsheng Li4 1 Sense Time Research 2 Nanyang Technological University 3 National University of Singapore 4 Multimedia Laboratory, The Chinese University of Hong Kong {renjiawei, zhaohaiyu, yishuai}@sensetime.com cunjun.yu@gmail.com shen0152@e.ntu.edu.sg xiao-ma@comp.nus.edu.sg hsli@ee.cuhk.edu.hk
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks that are clearly labeled or formatted like code.
Open Source Code Yes Corresponding author Code available at https://github.com/jiawei-ren/Balanced Meta Softmax
Open Datasets Yes We perform experiments on long-tailed image classification datasets, including CIFAR-10-LT [18], CIFAR-100-LT [18], Image Net-LT [23] and Places-LT [34] and one long-tailed instance segmentation dataset, LVIS [7].
Dataset Splits Yes For classification tasks, after training on the long-tailed dataset, we evaluate the models on the corresponding balanced test/validation dataset and report top-1 accuracy. We create the meta set by class-balanced sampling from the training set Dtrain. For LVIS, we use official training and testing splits.
Hardware Specification No The paper does not provide specific hardware details (e.g., exact GPU/CPU models, memory amounts) used for running its experiments.
Software Dependencies No The paper does not provide specific ancillary software details with version numbers (e.g., library or solver names with version numbers) needed to replicate the experiment.
Experiment Setup No Our implementation details can be found in the supplementary materials.