FoveaBox: Beyond Anchor-based Object Detector

By Tao Kong, Fuchun Sun, Huaping Liu, Yuning Jiang, Jianbo Shi. 2019

226 citations recorded by the source corpus.

We present FoveaBox, an accurate, flexible and completely anchor-free framework for object detection. While almost all state-of-the-art object detectors utilize the predefined anchors to enumerate possible locations, scales and aspect ratios for the search of the objects, their performance and generalization ability are also limited to the design of anchors. Instead, FoveaBox directly learns the object existing possibility and the bounding box coordinates without anchor reference. This is achieved by: (a) predicting category-sensitive semantic maps for the object existing possibility, and (b) producing category-agnostic bounding box for each position that potentially contains an object. The scales of target boxes are naturally associated with feature pyramid representations for each input image. Without bells and whistles, FoveaBox achieves state-of-the-art single model performance of 42.1 AP on the standard COCO detection benchmark. Specially for the objects with arbitrary aspect ratios, FoveaBox brings in significant improvement compared to the anchor-based detectors. More surprisingly, when it is challenged by the stretched testing images, FoveaBox shows great robustness and generalization ability to the changed distribution of bounding box shapes. The code will be made publicly available.

Subfields: Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

Citations in the corpus

Generalized Focal Loss: Learning Qualified and Distributed Bounding Boxes for Dense Object Detection (2020, methodology): “9 NeurIPS19 FoveaBox [13] R-101 18 X 13.”
Conditional DETR for Fast Training Convergence (2021, methodology): “Typical methods include YOLOv1 [30], CornerNet [21], ExtremeNet [50], CenterNet [49, 6], FCOS [39] and others [23, 28, 52, 19, 51, 22, 15, 46, 47].”
PP-YOLO: An Effective and Efficient Implementation of Object Detector (2020, methodology): “Anchor-point based detectors perform object bounding box regression based on anchor points instead of anchor boxes, including FSAF [49], FCOS[36], FoveaBox[17], SAPD[48].”
Dense Regression Network for Video Grounding (2020, background): “propose FoveaBox [24] to predict category-sensitive semantic maps for the object existing possibility and produce a category-agnostic bounding box at each position.”
VisDrone-DET2019: The Vision Meets Drone Object Detection in Image Challenge Results (2019, background): “On the other hand, Kong et al. [17] propose an accurate, ﬂexible and completely anchor-free framework, which pre-214 dicts category-sensitive semantic maps for the object existing possibility and category-agnostic bounding box for each position that potentially contains an object.”
PolarMask: Single Shot Instance Segmentation With Polar Representation (2019, background): “Recent trends in the vision community have spent more effort in designing simpler pipelines of bounding box detectors [17, 23, 29, 31, 33, 10, 19] and subsequent instance-wise recognition tasks including instance segmentation [2, 4, 34], which is also the main focus of our work here.”
FreeAnchor: Learning to Match Anchors for Visual Object Detection (2019, methodology): “It outperformed the baseline RetinaNet [7] and the anchor-free approaches including FoveaBox [22], FSAF [23], FCOS [13] and CornerNet [14].”
Seesaw Loss for Long-Tailed Instance Segmentation (2020, background): “Classical single-stage approaches [28, 31] require densely populated anchors as a prior, while anchor-free methods [24, 25, 45, 51] manage to achieve similar or better performance without such prior.”
Training-Time-Friendly Network for Real-Time Object Detection (2019, methodology): “To alleviate it, previous work either places object of different scale in different level by using FPN (Tian et al. 2019; Kong et al. 2019), or produces just one training sample based on the annotated box (Zhou, Wang, and Kr¨ahenb¨uhl 2019), as shown in Figure 3.”
Learning Spatial Fusion for Single-Shot Object Detection (2019, background): “Moreover, a number of studies [38, 45, 36, 14] indicate that the anchor-free pipeline contributes to considerably better performance with simpler designs.”