Search papers, labs, and topics across Lattice.
This paper addresses the challenge of model forgetting in long-incremental 3D object detection, where detectors must learn new object classes while retaining knowledge of previously learned ones. The authors identify a self-reinforcing cycle of data distribution shifts that exacerbate forgetting and propose the Learning-Dynamics-driven Memory and Review (LDMR) framework, which incorporates human-like review mechanisms and scene-aware memory evolution. Experimental results on indoor benchmarks demonstrate that LDMR significantly reduces forgetting and outperforms existing methods by a substantial margin.
Model forgetting in long-incremental 3D object detection can be effectively mitigated through human-inspired learning dynamics, leading to superior performance on challenging benchmarks.
Incremental 3D object detection requires a detector to learn novel object classes while remembering previously learned ones over sequentially arriving data. Previous methods, primarily based on pseudo-labeling, perform reasonably in short-incremental stages but still suffer from severe model forgetting when dealing with long-incremental sequences. We investigate this failure and reveal a detrimental self-reinforcing cycle: data distribution shift of novel classes causes model forgetting on old classes, which further produces accumulated error in pseudo-labeling that exacerbates model degradation. To address this issue, we draw inspiration from the human learning process and propose the \emph{Learning-Dynamics-driven Memory and Review} (LDMR) framework. LDMR monitors per-class detection quality at periodic training checkpoints and uses these learning-dynamics signals to drive two innovative mechanisms, namely (i) human-like intra-stage review that divides each incremental stage into multiple sub-stages'training and concentrates on remembering the most-forgotten objects, and (ii) scene-aware cross-stage memory evolution that evolves a memory bank to transfer knowledge between two consecutive stages by jointly considering scene learnability and diversity. Extensive experiments across multiple long-incremental protocols on indoor benchmarks SUN RGB-D and ScanNetV2 show that LDMR substantially mitigates the model forgetting and outperforms all baselines by a clear margin. Code is available at https://github.com/qianpeisheng/LDMR.
