Summary of Fred: Towards a Full Rotation-equivariance in Aerial Image Object Detection, by Chanho Lee et al.
FRED: Towards a Full Rotation-Equivariance in Aerial Image Object Detection
by Chanho Lee, Jinsu Son, Hyounguk Shon, Yunho Jeon, Junmo Kim
First submitted to arxiv on: 22 Dec 2023
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 The paper proposes a Fully Rotation-Equivariant Oriented Object Detector (FRED) that achieves end-to-end rotation-equivariance in oriented object detection. FRED decouples invariant and equivariant tasks to derive rotation-invariant features, representing bounding boxes as sets of rotation-equivariant vectors. These vectors are used as offsets in deformable convolution, enhancing spatial adaptation. The proposed method demonstrates higher robustness to image-level rotation compared to existing methods, achieving comparable performance on DOTA-v1.0 and outperforming by 1.5 mAP on DOTA-v1.5 while reducing model parameters to 16%. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary FRED is a new way of detecting objects that’s really good at dealing with pictures that are rotated. Most object detectors can handle pictures that are moved around, but they struggle when the picture is turned. FRED fixes this problem by making sure every step of the detection process is done in a way that doesn’t care about rotation. This makes it better than other methods at detecting objects in pictures that have been rotated. |
Keywords
* Artificial intelligence * Object detection
