RILOD: Near Real-Time Incremental Learning for Object Detection at the Edge

TL;DR

RILOD is a practical system enabling near real-time incremental learning for object detection on edge devices, allowing models to quickly adapt to new classes without forgetting old ones.

Contribution

The paper introduces a novel incremental learning algorithm for one-stage object detection models and a real-time dataset construction pipeline for edge applications.

Findings

Detects new object classes in minutes, including data collection and training.

Outperforms traditional fine-tuning in speed and reduces manual labeling.

Maintains detection accuracy for old classes while learning new ones.

Abstract

Object detection models shipped with camera-equipped edge devices cannot cover the objects of interest for every user. Therefore, the incremental learning capability is a critical feature for a robust and personalized object detection system that many applications would rely on. In this paper, we present an efficient yet practical system, RILOD, to incrementally train an existing object detection model such that it can detect new object classes without losing its capability to detect old classes. The key component of RILOD is a novel incremental learning algorithm that trains end-to-end for one-stage deep object detection models only using training data of new object classes. Specifically to avoid catastrophic forgetting, the algorithm distills three types of knowledge from the old model to mimic the old model's behavior on object classification, bounding box regression and feature extraction. In addition, since the training data for the new classes may not be available, a real-time dataset construction pipeline is designed to collect training images on-the-fly and automatically label the images with both category and bounding box annotations. We have implemented RILOD under both edge-cloud and edge-only setups. Experiment results show that the proposed system can learn to detect a new object class in just a few minutes, including both dataset construction and model training. In comparison, traditional fine-tuning based method may take a few hours for training, and in most cases would also need a tedious and costly manual dataset labeling step.