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.

SSIT: a Sample Selection-Based Incremental Model Training Method for Image Recognition

A Novel Approach to Incremental SVM Learning Algorithm

An Incremental Updating Method for Support Vector Machines

IKNN-SVM: A Hybrid Incremental Algorithm for Image Classification

Few-Shot Incremental Learning for Label-to-Image Translation

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

Class-Incremental Learning: Survey and Performance Evaluation on Image Classification

An integrated classification model for incremental learning

An Incremental Learning Algorithm Based on Support Vector Machine for Pattern Recognition

Class-Incremental Recognition of Objects in Remote Sensing Images with Dynamic Hybrid Exemplar Selection

Preliminary Exploration of Data Incremental Learning Method

An incremental structured part model for image classification

New Incremental Learning Algorithm with Support Vector Machines

An incremental learning face recognition system for single sample per person

A Practical Incremental Method to Train Deep CTR Models

Effective discriminative TCM-KNN for incremental learning

Image segmentation application based on incremental learning vector SVM algorithm

An incremental learning method based on SVM for online sketchy shape recognition

Optimized incremental SVM algorithm based on support vector threshold control

Dynamic Incremental Learning in Medical Imaging

A Gradual Training Algorithm of Incremental Support Vector Machine Learning