ACE: Anchor-Free Corner Evolution for Real-Time Arbitrarily-Oriented Object Detection.

Area of Science:

• Computer Vision
• Machine Learning
• Image Analysis

Background:

• Axis-aligned bounding boxes are insufficient for compactly enclosing objects with arbitrary orientations in real-world scenarios.
• Arbitrarily-oriented object detection is a growing area of research due to the limitations of standard bounding boxes.

Purpose of the Study:

• To propose a novel and effective model for detecting arbitrarily-oriented objects.
• To improve upon existing methods by avoiding direct angle prediction for oriented bounding boxes.

Main Methods:

• The model evolves axis-aligned bounding boxes into oriented quadrilateral boxes using dynamically gathered contour information.
• Key points are established on sampled contour points of the axis-aligned bounding box.
• A dynamic information gathering mechanism enriches key point features by propagating geometrical and semantic information.

Main Results:

• The proposed model demonstrates effectiveness and advantage across multiple datasets, including aerial imagery (HRSC2016, DOTA) and scene text/hand datasets (ICDAR2015, TD500, Oxford-Hand).
• Experimental results validate the model's capability in accurately detecting objects with varying orientations.

Conclusions:

• The novel approach of evolving bounding boxes with dynamic contour information gathering offers a significant advancement in arbitrarily-oriented object detection.
• The method provides a more compact and accurate representation for oriented objects compared to traditional axis-aligned bounding boxes.