Abstract:Fine-grained remote sensing object detection aims at precisely locating objects and determining the fine-level categories. This task is exceptionally challenging due to the substantial interclass similarity, presenting difficulties in capturing discriminative features. We attribute this to the absence of essential information that can serve as supervision for the learning. This involves comprehensive visual patterns of objects and intrinsic relationships of multigranular features. In this article, we propose a novel scheme dubbed as digging into the multigranular archetype (DIMA) for fine-grained remote sensing object detection. In detail, we first design a simple yet effective frequency-aware representation supplement (FARS) mechanism learning from original images and their auxiliary frequency counterparts simultaneously. The FARS introduces high- and low-frequency representations to reinforce a range of visual cues, such as particular regions associated with the former and contours of objects related to the latter. Then, we further devise a module named hierarchical classification paradigm (HCP), which constructs the interhierarchy relationships between coarse and fine-level representations and then exploits them to guide fine-grained feature enhancement. HCP eventually selects and boosts samples that are hard to discriminate by keeping consistency in multilevels. Our method can be easily integrated into prevailing oriented object detectors and brings consistent performance improvements across these detectors. Notably, our method combined with oriented RCNN (ORCNN) achieves 44.44% (+3.62%) on the FAIR1M and 91.0% (+6.9%) on the MAR20. Moreover, thoughtful discussions about qualitative results and rich visualizations are provided to intuitively underscore the superiority of our approach. The source code is available at https://github.com/chengjc2019/DIMA.

Adaptive Contour Features in Oriented Granular Space for Human Detection and Segmentation

Granularity-tunable Gradients Partition (GGP) Descriptors for Human Detection.

Human Detection Based on Fusion of Histograms of Oriented Gradients and Main Partial Features

Real-time human detection based on gentle MILBoost with variable granularity HOG-CSLBP

An HOG-CT Human Detector with Histogram-Based Search.

AdaContour: Adaptive Contour Descriptor with Hierarchical Representation

Neighborhood Aided Implicit Active Contours

Learning Sparse Features in Granular Space for Multi-View Face Detection

Contour-motion Feature (CMF): A Space-Time Approach for Robust Pedestrian Detection

Boosting Associated Pairing Comparison Features for Pedestrian Detection.

Oriented Gradients for Human Action Recognition.

Contour Segment Grouping for Object Detection.

A feature aggregation network for contour detection inspired by complex cells properties

DIMA: Digging into Multigranular Archetype for Fine-Grained Object Detection

Real-time Human Detection Using Contour Cues.

A novel hierarchical framework for human head-shoulder detection

Robust Domain Adaptive Object Detection with Unified Multi-Granularity Alignment

Convex-hull Feature Adaptation for Oriented and Densely Packed Object Detection

GRA: Detecting Oriented Objects through Group-wise Rotating and Attention

Anchor-Free Oriented Proposal Generator for Object Detection

Scale adaption-guided human face detection