Abstract:With the rapid development of remote sensing technology and the growing demand for applications, the classical deep learning-based object detection model is bottlenecked in processing incremental data, especially in the increasing classes of detected objects. It requires models to sequentially learn new classes of objects based on the current model, while preserving old categories-related knowledge. Existing class-incremental detection methods achieve this goal mainly by constraining the optimization trajectory in the feature of output space. However, these works neglect the case where the previously learned background is a new category to learn, resulting in performance degradation in the new category because of the conflict between remaining the background-related knowledge or updating the background-related knowledge. This paper proposes a novel class-incremental detection method incorporated with the teacher-student architecture and the selective distillation (SDCID) strategy. Specifically, it is the asymmetry architecture, i.e., the teacher network temporarily stores historical knowledge of previously learned objects, and the student network integrates historical knowledge from the teacher network with the newly learned object-related knowledge, respectively. This asymmetry architecture reveals the significance of the distinct representation of history knowledge and new knowledge in incremental detection. Furthermore, SDCID selectively masks the shared subobject of new images to learn and previously learned background, while learning new categories of images and then transfers the classification results of the student model to the background class following the judgment model of the teacher model. This manner avoids interferences between the background category-related knowledge from a teacher model and the learning of other new classes of objects. In addition, we proposed a new incremental learning evaluation metric, C-SP, to comprehensively evaluate the incremental learning stability and plasticity performance. We verified the proposed method on two object detection datasets of remote sensing images, i.e., DIOR and DOTA. The experience results in accuracy and C-SP suggest that the proposed method surpasses existing class-incremental detection methods. We further analyzed the influence of the mask component in our method and the hyper-parameters sensitive to our method.

Incremental Learning of Boosted Face Detector

Incremental Training of a Detector Using Online Sparse Eigendecomposition

Multiclass AdaBoost ELM and Its Application in LBP Based Face Recognition

Improving Part based Object Detection by Unsupervised, Online Boosting

Z-Adaboost: Boosting 2-thresholded weak classifiers for object detection

Two-Stage Multi-Class Adaboost for Facial Expression Recognition

Incremental Learning In Online Scenario

Towards Generalized and Incremental Few-Shot Object Detection

Online Boosting OC for Face Recognition in Continuous Video Stream

Real-time face detection with self-adaptive cost-sensitive AdaBoost

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

On the Design of Cascades of Boosted Ensembles for Face Detection

Incremental Detection of Remote Sensing Objects with Feature Pyramid and Knowledge Distillation

Modification of the AdaBoost-based Detector for Partially Occluded Faces

Incremental Learning of Object Detectors without Catastrophic Forgetting

Efficiently Training a Better Visual Detector with Sparse Eigenvectors

Boosted pedestrian detector adaptation in specific scenes

LACBoost and FisherBoost: Optimally Building Cascade Classifiers

Data level object detector adaptation with online multiple instance samples

A Class-Incremental Detection Method of Remote Sensing Images Based on Selective Distillation

Efficiently Learning a Detection Cascade with Sparse Eigenvectors