Annotator: A Generic Active Learning Baseline for LiDAR Semantic Segmentation

Authors: Binhui Xie, Shuang Li, Qingju Guo, Chi Liu, Xinjing Cheng

NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In this section, we conduct extensive experiments on several public benchmarks under three active learning scenarios: (i) AL setting where all available data points are from unlabeled target domain; (ii) ASFDA setting where we can only access a pre-trained model from the source domain; (iii) ADA setting where all data points from source domain can be utilized and a portion of unlabeled target data is selected to be annotated. We first introduce the dataset used in this work and experimental setup and then present experimental results of baseline methods and extensive analyses of Annotator.
Researcher Affiliation Collaboration Binhui Xie Beijing Institute of Technology binhuixie@bit.edu.cn Shuang Li Beijing Institute of Technology shuangli@bit.edu.cn Qingju Guo Beijing Institute of Technology qingjuguo@bit.edu.cn Chi Harold Liu Beijing Institute of Technology chiliu@bit.edu.cn Xinjing Cheng Tsinghua University & Inceptio Technology cnorbot@gmail.com
Pseudocode No The paper describes methods in text but does not contain structured pseudocode or algorithm blocks clearly labeled as such.
Open Source Code No Project page: https://binhuixie.github.io/annotator-web/ (As of the current date, the project page states 'Code coming soon...')
Open Datasets Yes We build all benchmarks upon Syn Li DAR [84], Semantic KITTI [3], Semantic POSS [42], and nu Scenes [4], constructing two simulation-to-real and two real-to-real adaptation scenarios.
Dataset Splits Yes Semantic KITTI (KITTI) [3] is a popular Li DAR segmentation dataset, including 2,9130 training scans and 6,019 validation scans with 19 categories. Semantic POSS (POSS) [42] consists of 2,988 real-world scans with point-level annotations over 14 semantic classes. As suggested in [42], we use the sequence 03 for validation and the remaining sequences for training.
Hardware Specification Yes all experiments share the same backbones and are within the same codebase, which are implemented using Py Torch [43] on a single NVIDIA Tesla A100 GPU.
Software Dependencies No The paper mentions 'implemented using Py Torch [43]' but does not specify a version number for PyTorch or any other software libraries or dependencies. Therefore, a reproducible description with specific version numbers is not provided.
Experiment Setup Yes We use the SGD optimizer and adopt a cosine learning rate decay schedule with initial learning rate of 0.01. And the batch size for both source and target data is 16. For additional details, please consult Appendix A.2.