Learnable Loss Balancing in Anchor-Free Oriented Detectors for Aerial Object

Article Geochemistry & Geophysics

Guiding Clean Features for Object Detection in Remote Sensing Images

Gong Cheng et al.

Summary: Object detection in remote sensing images has made significant progress, but it faces two defects related to feature pyramid and traditional label assignment strategy. A new framework proposed by the authors tackles these issues effectively with two key components AFPN and GAS.

IEEE GEOSCIENCE AND REMOTE SENSING LETTERS (2022)

Add to Collection

Article Computer Science, Artificial Intelligence

A General Gaussian Heatmap Label Assignment for Arbitrary-Oriented Object Detection

Zhanchao Huang et al.

Summary: A novel arbitrary-oriented object detection (AOOD) method called General Gaussian Heatmap Label Assignment (GGHL) is proposed in this paper to improve AOOD performance. This method utilizes a new label assignment strategy and an oriented-bounding-box representation component to adapt to the characteristics of different objects.

IEEE TRANSACTIONS ON IMAGE PROCESSING (2022)

Add to Collection

Proceedings Paper Computer Science, Artificial Intelligence

RFLA: Gaussian Receptive Field Based Label Assignment for Tiny Object Detection

Chang Xu et al.

Summary: This paper proposes a Gaussian Receptive Field based Label Assignment (RFLA) strategy and a Hierarchical Label Assignment (HLA) module for tiny object detection to address the sub-optimality issue of current methods. Extensive experiments demonstrate the effectiveness of the proposed methods in achieving balanced learning and outperforming state-of-the-art competitors.

COMPUTER VISION, ECCV 2022, PT IX (2022)

Add to Collection

Article Geochemistry & Geophysics

Object Detection for Aerial Images With Feature Enhancement and Soft Label Assignment

Ying Yu et al.

Summary: In this article, an anchor-free detector with feature enhancement and soft label assignment (FSDet) is introduced for object detection in aerial images. It addresses the challenges of arbitrary-oriented objects and imbalanced foreground-background distribution. The proposed method achieves competitive performance by aligning features with oriented objects, suppressing background context, and assigning soft labels to optimize training.

IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING (2022)

Add to Collection

Article Environmental Sciences

A2S-Det: Efficiency Anchor Matching in Aerial Image Oriented Object Detection

Zhifeng Xiao et al.

Summary: The study introduces a detection method that adapts anchoring based on sample balance, selecting candidate anchors through horizontal IoU and using an adaptive threshold module to maintain a balance between positive and negative anchors, ultimately achieving better performance on a public aerial image dataset.

REMOTE SENSING (2021)

Add to Collection

Article Geochemistry & Geophysics

Learning Center Probability Map for Detecting Objects in Aerial Images

Jinwang Wang et al.

Summary: This article proposes to convert the OBB regression problem into a CenterMap prediction problem, which largely eliminates ambiguities and improves efficiency and accuracy.

IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING (2021)

Add to Collection

Article Geochemistry & Geophysics

GRS-Det: An Anchor-Free Rotation Ship Detector Based on Gaussian-Mask in Remote Sensing Images

Xiangrong Zhang et al.

Summary: An anchor-free rotation ship detector, named GRS-Det, is proposed in this paper to address the ship detection challenges with rotation boxes. By utilizing a feature extraction network with selective concatenation module and a rotation Gaussian-Mask model, the detector achieves state-of-the-art results on ship detection benchmarks.

IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING (2021)

Add to Collection

Article Computer Science, Artificial Intelligence

Gliding Vertex on the Horizontal Bounding Box for Multi-Oriented Object Detection

Yongchao Xu et al.

Summary: A framework for detecting multi-oriented objects is proposed in this paper, which accurately describes multi-oriented objects by sliding bounding box vertices and introducing additional target variables, achieving superior performances on multiple object detection benchmarks.

IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE (2021)

Add to Collection

Article Environmental Sciences