Single-Image 3D Scene Parsing Using Geometric Commonsense

Authors: Chengcheng Yu, Xiaobai Liu, Song-Chun Zhu

IJCAI 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Results with comparisons on public datasets showed that our method clearly outperforms the alternative methods.
Researcher Affiliation Academia 1 Department of Statistics, University of California, Los Angeles (UCLA), CA 2 Department of Computer Science, San Diego State Univesity (SDSU), CA chengchengyu@ucla.edu, xiaobai.liu@mail.sdsu.edu, sczhu@stat.ucla.edu
Pseudocode No The paper describes the inference process but does not include structured pseudocode or an algorithm block.
Open Source Code No The paper does not provide any concrete statement or link regarding the availability of its source code.
Open Datasets Yes To evaluate the proposed method, we collect an image dataset that covers five different categories: 1) country; 2) suburban; 3) road; 4) campus; and 5) urban, resulting in a collection of 2,500 images. The urban images mostly follow the Manhattan assumption [Liu et al., 2014] while the country images do not. These images are selected from existing datasets [Hoiem et al., 2008; Liu et al., 2014; Everingham et al., 2015].
Dataset Splits No The paper mentions 'For every category, we use 100 images for training and use the rest for testing.', but it does not specify a separate validation split or explicit cross-validation.
Hardware Specification Yes It costs about 1-2 minutes to converge on a Dell Workstation (i7-4790 CPU@3.6GHZ with 16GB memory).
Software Dependencies No The paper mentions various methods used (e.g., 'method by Ren et al.', 'method by Li et al.'), but it does not specify software names with version numbers.
Experiment Setup Yes We set the maximal iteration of DDMCMC to be 2000.