Abstract:Object detection has made tremendous strides in computer vision. Small object detection with appearance degradation is a prominent challenge, especially for aerial observations. To collect sufficient positive/negative samples for heuristic training, most object detectors preset region anchors in order to calculate intersection-over-union (IoU) against the ground-truth data. In this case, small objects are frequently abandoned or mislabeled. In this article, we present an effective dynamic enhancement anchor network (DEA-Net) to construct a novel training sample generator. Different from the other state-of-the-art (SOTA) techniques, the proposed network leverages a sample discriminator to realize interactive sample screening between an anchor-based unit and an anchor-free unit to generate eligible samples. Besides, multi-task joint training with a conservative anchor-based inference scheme enhances the performance of the proposed model while reducing computational complexity. The proposed scheme supports both oriented and horizontal object detection tasks. Extensive experiments on two challenging aerial benchmarks (i.e., Dataset of Object deTection in Aerial images (DOTA) and HRSC2016) indicate that our method achieves SOTA performance in accuracy with moderate inference speed and computational overhead for training. On DOTA, our DEA-Net which integrated with the baseline of RoI-transformer surpasses the advanced method by 0.40% mean-average-precision (mAP) for oriented object detection with a weaker backbone network (ResNet-101 vs. ResNet-152) and 3.08% mAP for horizontal object detection with the same backbone. Besides, our DEA-Net which integrated with the baseline of ReDet achieves the SOTA performance by 80.37%. On HRSC2016, it surpasses the previous best model by 1.1% using only three horizontal anchors. The source code and the training set are made publicly available at https://github.com/QxGeng/DEA-Net.

Hard Anchor Attention in Anchor-based Detector

Bridging the Gap Between Anchor-based and Anchor-free Detection via Adaptive Training Sample Selection

Effective Anchor Adaptation and Feature Enhancement Strategies for Tiny Object Detection in Aerial Images

Learning from Noisy Anchors for One-Stage Object Detection.

MetaAnchor: Learning to Detect Objects with Customized Anchors.

Anchor Retouching via Model Interaction for Robust Object Detection in Aerial Images

Augmenting Anchors by the Detector Itself.

HAMBox: Delving into Online High-quality Anchors Mining for Detecting Outer Faces

Dynamic Anchor Learning for Arbitrary-Oriented Object Detection

PosNeg-Balanced Anchors with Aligned Features for Single-Shot Object Detection

AABO: Adaptive Anchor Box Optimization for Object Detection via Bayesian Sub-sampling

Improving Object Detection with Inverted Attention

Anchor Box Optimization for Object Detection

Revisiting Foreground-Background Imbalance in Object Detectors

Are Sampling Heuristics Necessary in Object Detectors

Object as Hotspots: An Anchor-Free 3D Object Detection Approach via Firing of Hotspots

SAFNet: A Semi-Anchor-Free Network with Enhanced Feature Pyramid for Object Detection.

Is Heuristic Sampling Necessary in Training Deep Object Detectors?

Pixel-Anchor: A Fast Oriented Scene Text Detector with Combined Networks

Multiple Anchor Learning for Visual Object Detection

Center-Boundary Dual Attention for Oriented Object Detection in Remote Sensing Images