Renovating Names in Open-Vocabulary Segmentation Benchmarks

Authors: Haiwen Huang, Songyou Peng, Dan Zhang, Andreas Geiger

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

Reproducibility Variable Result LLM Response
Research Type Experimental 5 Experiments
Researcher Affiliation Collaboration 1 Bosch Io C Lab, University of T ubingen 3 T ubingen AI Center 3 Autonomous Vision Group, University of T ubingen 4 ETH Zurich 5 MPI for Intelligent Systems, T ubingen 6 Bosch Center for Artificial Intelligence
Pseudocode No The paper describes the method using text and figures (e.g., Fig. 2), but does not contain a formal pseudocode or algorithm block.
Open Source Code Yes We provide our code and relabelings for several popular segmentation datasets to the research community on our project page: https://andrehuang.github.io/renovate/ .
Open Datasets Yes We renovate three panoptic segmentation datasets respectively: MS COCO [13], ADE20K [14], and Cityscapes [15].
Dataset Splits Yes We train a renaming model for 60k iterations with a batch size of 16 on the training set, then generate RENOVATE names for the entire dataset.
Hardware Specification Yes Training one renaming model on 4 A-100 GPUs requires approximately 3 days.
Software Dependencies No The paper mentions software like GPT-4, CLIP, Mask2Former, CaSED, and AdamW, but does not provide specific version numbers for these or other key software components used in the experiments.
Experiment Setup Yes We train a renaming model for 60k iterations with a batch size of 16 on the training set