Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

SID: Incremental Learning for Anchor-Free Object Detection via Selective and Inter-Related Distillation

About

Incremental learning requires a model to continually learn new tasks from streaming data. However, traditional fine-tuning of a well-trained deep neural network on a new task will dramatically degrade performance on the old task -- a problem known as catastrophic forgetting. In this paper, we address this issue in the context of anchor-free object detection, which is a new trend in computer vision as it is simple, fast, and flexible. Simply adapting current incremental learning strategies fails on these anchor-free detectors due to lack of consideration of their specific model structures. To deal with the challenges of incremental learning on anchor-free object detectors, we propose a novel incremental learning paradigm called Selective and Inter-related Distillation (SID). In addition, a novel evaluation metric is proposed to better assess the performance of detectors under incremental learning conditions. By selective distilling at the proper locations and further transferring additional instance relation knowledge, our method demonstrates significant advantages on the benchmark datasets PASCAL VOC and COCO.

Can Peng, Kun Zhao, Sam Maksoud, Meng Li, Brian C. Lovell• 2020

Related benchmarks

TaskDatasetResultRank
Object DetectionMS COCO 2017 (minival)
AP34
50
Class Incremental Object DetectionCOCO 2017 (val)
AP45.7
30
Object DetectionCOCO 40+40 scenario 2017 (val)
AP34
25
Object DetectionCOCO 70+10 scenario 2017 (val)
AP32.8
25
Incremental Object DetectionCOCO 2017 (40-40)
AP34
19
Incremental Object DetectionCOCO 2017 (70-10)
AP32.8
19
Incremental Object DetectionCOCO 40-10 multi-step--
13
Incremental Object DetectionCOCO 40-20 x 2 2017
Score (60-80)23.8
9
Object DetectionCOCO (last 40 classes + first 40 classes)
AP33.5
6
Incremental Object DetectionCOCO classes 1-60, incremental step 40-60
AP34
4
Showing 10 of 15 rows

Other info

Follow for update