Amodal Instance Segmentation via Prior-Guided Expansion

Authors: Junjie Chen, Li Niu, Jianfu Zhang, Jianlou Si, Chen Qian, Liqing Zhang

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

Reproducibility Variable Result LLM Response
Research Type Experimental We conduct extensive experiments on KINS, D2SA, and COCOA cls datasets, which show the effectiveness of our method.
Researcher Affiliation Collaboration Junjie Chen1, Li Niu1*, Jianfu Zhang1, Jianlou Si2, Chen Qian2, Liqing Zhang1* 1 The Mo E Key Lab of AI, CSE department, Shanghai Jiao Tong University 2 Sense Time Research, Sense Time
Pseudocode No The paper describes the methodology in text and figures, but does not include structured pseudocode or algorithm blocks.
Open Source Code No The paper states 'We implement the proposed method on the codebase of previous work (Xiao et al. 2021)', but it does not provide an explicit statement about releasing its own source code or a link to a repository.
Open Datasets Yes We conduct extensive experiments on three datasets: KINS (Qi et al. 2019), D2SA (Follmann et al. 2019), and COCOA cls (Zhu et al. 2017).
Dataset Splits No The paper uses training instances and mentions training objectives, but it does not specify explicit train/validation/test dataset splits with percentages, counts, or references to predefined splits.
Hardware Specification Yes We conducted the experiments on Ubuntu 18.04 system with 32 GB Intel 9700K CPU and two NVIDIA 1080ti GPU cards.
Software Dependencies Yes We implement the proposed method on the codebase of previous work (Xiao et al. 2021), which builds on Detectron2 using Python 3.7 and Py Torch 1.4.0 framework.
Experiment Setup Yes For Mask R-CNN, we follow the setup in previous works (Xiao et al. 2021; Follmann et al. 2019), e.g., predicting amodal box in its box head, using Res Net-50 (He et al. 2016) as backbone, and using channel size C = 256 and spatial size 14 x 14 for the region feature maps of mask heads.