Abstract:Remote sensing images (RSIs) frequently contain quite a few tiny objects with a finite number of pixels to study. The limited spatial information poses a challenge for extracting discriminative features for representing the characteristics of tiny objects. Existing solutions mainly focus on aggregating contextual information at different levels, while rarely touching the step that is crucial for model training, i.e., label assignment. Tiny instances occupy fairly small regions of images and have limited overlaps to priors (anchors or dots), which is a dilemma for traditional label assignment strategies. Despite being simple and effective, the mainstream Intersection over Union (IoU)-based label assignment strategy struggles to accurately measure the localization of tiny bounding boxes. In contrast, the Kullback–Leibler divergence (KLD) localization metric accurately reflects minor offsets of tiny bounding boxes. More importantly, KLD is able to measure non-overlapping bounding boxes, providing an advantage in mining more potential positive samples of tiny objects. In this article, from a cost-efficient point of view, we detect tiny objects through KLD in the form of single-stage framework. Specifically, we model the parameterized bounding box as a 2-D Gaussian distribution (Bbox2Gaussian) in order to use KLD as a localization metric. Then, we propose an adaptive online training sample mining (Ali-TSM) strategy based on inter-distribution similarity, which selects high-quality positive samples by considering localization and classification rather than just centroid distance or IoU. Finally, task-level attention (TlA) is introduced to guide the model in freely selecting the appropriate features for the classification or regression task. We conducted extensive experiments on four popular public datasets. Compared to the baseline, KLDet improves performance on Tiny Object Detection in Aerial Images (AI-TOD) and object DetectIon in Optical Remote sensing image (DIOR) by 4.1 AP and 6.7 mAP. On VisDrone and Small Object Detection dAtasets (SODA-D), KLDet exhibits superior performance than baseline. The code is available at https://github.com/TinyOD/mmdet-kldet.

Multibox Sample Selection for Active Object Detection.

Dive Deeper into Box for Object Detection

Pbde: An Effective Method For Filtering False Positive Boxes In Object Detection

PBDE: an Effective Post-Processing Method Based on Box Density for Object Detection

Box-Level Active Detection

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

Spatial Self-Distillation for Object Detection with Inaccurate Bounding Boxes

DuBox: No-Prior Box Objection Detection via Residual Dual Scale Detectors

Multiple instance active learning for object detection

Detection Selection Algorithm: A Likelihood based Optimization Method to Perform Post Processing for Object Detection

Vaccination of pediatric nurses with live attenuated cytomegalovirus.

Classification Committee for Active Deep Object Detection

MUS-CDB: Mixed Uncertainty Sampling with Class Distribution Balancing for Active Annotation in Aerial Object Detection

SSD: Single Shot MultiBox Detector

Scalable, High-Quality Object Detection

Correction: Structure-Activity Relationships of Constrained Phenylethylamine Ligands for the Serotonin 5-HT2 Receptors

MultiOOD: Scaling Out-of-Distribution Detection for Multiple Modalities

KLDet: Detecting Tiny Objects in Remote Sensing Images via Kullback–Leibler Divergence

Online Pseudo-Label Unified Object Detection for Multiple Datasets Training

IoU_MDA: An Occluded Object Detection Algorithm Based on Fuzzy Sample Anchor Box IoU Matching Degree Deviation Aware

Box-Level Active Detection – Supplementary Material