Summary of Inspacetype: Dataset and Benchmark For Reconsidering Cross-space Type Performance in Indoor Monocular Depth, by Cho-ying Wu et al.
InSpaceType: Dataset and Benchmark for Reconsidering Cross-Space Type Performance in Indoor Monocular Depth
by Cho-Ying Wu, Quankai Gao, Chin-Cheng Hsu, Te-Lin Wu, Jing-Wen Chen, Ulrich Neumann
First submitted to arxiv on: 25 Aug 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Machine Learning (cs.LG)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary In this paper, researchers investigate the common factor of space type in indoor monocular depth estimation, a crucial aspect for home automation and applications like robot navigation or AR/VR. Most previous methods focus on the NYUv2 Dataset and overall performance, but their generalization to diverse spaces has yet to be studied. The authors introduce the InSpaceType Dataset, a high-quality RGBD dataset for general indoor scenes, and benchmark 13 recent state-of-the-art methods on it. Their analysis shows that most methods suffer from performance imbalance between head and tailed types, with some top methods being more severe. The paper also extends the analysis to four datasets, discusses best practices in synthetic data curation, and conducts dataset ablation to identify key factors in generalization. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Indoor monocular depth estimation helps home automation. Most previous methods only test on one dataset and focus on overall performance. But what if they’re not good at other types of spaces? This paper looks into this common but overlooked factor: space type. They create a new dataset, InSpaceType, with lots of indoor scenes, and test 13 top methods on it. Most methods are bad at some space types, like head or tailed spaces. The authors also show how to make synthetic data for training models and what factors affect generalization. |
Keywords
» Artificial intelligence » Depth estimation » Generalization » Synthetic data
