Revisiting Open-Set Panoptic Segmentation
Authors: Yufei Yin, Hao Chen, Wengang Zhou, Jiajun Deng, Haiming Xu, Houqiang Li
AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experimental results on different datasets validate the effectiveness of our method. |
| Researcher Affiliation | Academia | 1 CAS Key Laboratory of Technology in GIPAS, EEIS Department, University of Science and Technology of China 2 Zhejiang University 3 Institute of Artificial Intelligence, Hefei Comprehensive National Science Center 4 Australian Institute for Machine Learning, University of Adelaide |
| Pseudocode | No | The paper describes algorithms (e.g., Unknown Segment Mining algorithm) but does not present them in a structured pseudocode or algorithm block. |
| Open Source Code | No | The paper does not contain any explicit statement or link indicating that the source code for the methodology is openly available. |
| Open Datasets | Yes | To involve more diverse categories and complete annotations, we construct a new LVIS-PS dataset for the OPS task based on the LVIS dataset (Gupta, Dollar, and Girshick 2019) and COCO. |
| Dataset Splits | Yes | We adopt the LVIS-PS dataset for the OPS task. We use the corresponding original COCO annotations during training, while using generated LVIS-PS annotations for inference. ... We evaluate our method on the proposed LVIS-PS dataset. During training, as discussed in Sec. 3.2, we use the corresponding original COCO annotations of LVIS-PS train set, which contain 80 thing classes and 53 stuff classes. LVIS-PS val set is utilized for evaluation, which has 994 classes in total. |
| Hardware Specification | No | This work was supported in part by the GPU cluster built by MCC Lab of Information Science and Technology Institution and the Supercomputing Center of the USTC. This statement is too general and does not specify particular GPU or CPU models, memory, or other hardware details. |
| Software Dependencies | No | The paper does not provide specific version numbers for any software dependencies, such as libraries or frameworks (e.g., PyTorch, TensorFlow, Python version). |
| Experiment Setup | Yes | For hyperparameters, the overlap thresholds are set to 0.8 and 0.9 for void and stuff areas, respectively. The score threshold for stuff areas is set to 0.3 in Semi-Pano FCN-2s. k is set to 50, which is the same with (Li et al. 2021). |