Abstract:Deep-learning-based object detectors have substantially improved state-of-the-art object detection in remote sensing images in terms of precision and degree of automation. Nevertheless, the large variation of the object scales makes it difficult to achieve high-quality detection across multiresolution remote sensing images, where the quality is defined by the Intersection over Union (IoU) threshold used in training. In addition, the imbalance between the positive and negative samples across multiresolution images worsens the detection precision. Recently, it was found that a Cascade region-based convolutional neural network (R-CNN) can potentially achieve a higher quality of detection by introducing a cascaded three-stage structure using progressively improved IoU thresholds. However, the performance of Cascade R-CNN degraded when the fourth stage was added. We investigated the cause and found that the mismatch between the ROI features and the classifier could be responsible for the degradation of performance. Herein, we propose a Cascade R-CNN++ structure to address this issue and extend the three-stage architecture to multiple stages for general use. Specifically, for cascaded classification, we propose a new ensemble strategy for the classifier and region of interest (RoI) features to improve classification accuracy at inference. In localization, we modified the loss function of the bounding box regressor to obtain higher sensitivity around zero. Experiments on the DOTA dataset demonstrated that Cascade R-CNN++ outperforms Cascade R-CNN in terms of precision and detection quality. We conducted further analysis on multiresolution remote sensing images to verify model transferability across different object scales.

Training a Multi-Exit Cascade with Linear Asymmetric Classification for Efficient Object Detection

Training Effective Node Classifiers for Cascade Classification

Optimally Training a Cascade Classifier

Efficiently Learning a Detection Cascade with Sparse Eigenvectors

LACBoost and FisherBoost: Optimally Building Cascade Classifiers

Linear Asymmetric Classifier for cascade detectors

Boosting 2-Thresholded Weak Classifiers over Scattered Rectangle Features for Object Detection.

Boosted Cascade of Scattered Rectangle Features for Object Detection.

Cascaded Iterative Training Model and Parallel Multi-Classifiers for Visual Object Tracking

Fast Asymmetric Learning for Cascade Face Detection

Asymmetric Pruning for Learning Cascade Detectors

CascadeV-Det: Cascade Point Voting for 3D Object Detection

Efficiently Training a Better Visual Detector with Sparse Eigenvectors

High Quality Object Detection for Multiresolution Remote Sensing Imagery Using Cascaded Multi-Stage Detectors.

Fast and Robust Object Detection Using Asymmetric Totally Corrective Boosting

On the Design of Cascades of Boosted Ensembles for Face Detection

Cascade R-CNN: Delving into High Quality Object Detection

Asymmetric Totally-Corrective Boosting for Real-Time Object Detection

Submodular Asymmetric Feature Selection in Cascade Object Detection.

Semi-Supervised and Long-Tailed Object Detection with CascadeMatch