Summary of Quantifying Context Bias in Domain Adaptation For Object Detection, by Hojun Son and Arpan Kusari
Quantifying Context Bias in Domain Adaptation for Object Detection
by Hojun Son, Arpan Kusari
First submitted to arxiv on: 23 Sep 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI); Robotics (cs.RO)
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 The paper explores domain adaptation for object detection, specifically studying context bias in transferring models between different domains. Various methods exist to minimize context bias, but this work focuses on analyzing changes in background features during adaptation and how context bias is represented in different domains. The authors experiment with varying activation values over different layers of trained models and masking the background to impact detection quality. They use three datasets (CARLA, Cityscapes, and Cityscapes foggy) to quantify context bias using metrics like Maximum Mean Discrepancy (MMD) and Maximum Variance Discrepancy (MVD). The findings suggest that understanding context bias can affect domain adaptation approach and focus. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Domain adaptation for object detection helps models learn to recognize objects in new environments. This paper looks at how well models adapt when the background changes, like from clear skies to foggy conditions. Researchers tried different ways to make this happen, such as changing what parts of the model get activated or hiding certain parts of the image. They used three kinds of data: one made-up set and two real sets (Cityscapes and Cityscapes foggy). By analyzing these datasets, they found that how well models adapt depends on how much context bias there is in the data. |
Keywords
» Artificial intelligence » Domain adaptation » Object detection
