SID: Incremental learning for anchor-free object detection via Selective and Inter-related Distillation

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

doi:10.1016/j.cviu.2021.103229

Cited by 27 publications

(22 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In addition, we also compared our method with LwF (Li & Hoiem, 2018), RILOD , and SID (Peng et al, 2021). Both Table 1 and Table 2 show that although LwF works well in incremental classification, it is even lower AP than directly fine-tuning in detection tasks.…”

Section: Experiments and Discussionmentioning

confidence: 99%

“…Thereafter, some researchers move this area forward. Peng et al (2021) proposed SID approach for incremental object detection on anchor-free detector and conducted experiments on FCOS (Tian et al, 2019) and CenterNet (Zhou et al, 2019). Li et al (2021a) studied object detection based on class-incremental learning on Faster RCNN detector with emphasis given to few-shot scenarios, which is also the focus of ONCE algorithm (Pérez-Rúa et al, 2020).…”

Section: Related Workmentioning

confidence: 99%

“…A straightforward way in incremental object detection is based on knowledge distillation (Hinton et al, 2015). Peng et al (2021) stressed that the Tower layers could reduce catastrophic forgetting significantly. They implemented incremental learning on an anchor-free detector and selectively performed distillation on non-regression outputs.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Response-based Distillation for Incremental Object Detection

Feng¹,

Wang²

2021

Preprint

View full text Add to dashboard Cite

Traditional object detection are ill-equipped for incremental learning. However, fine-tuning directly on a well-trained detection model with only new data will leads to catastrophic forgetting. Knowledge distillation is a straightforward way to mitigate catastrophic forgetting. In Incremental Object Detection (IOD), previous work mainly focuses on feature-level knowledge distillation, but the different response of detector has not been fully explored yet. In this paper, we propose a fully response-based incremental distillation method focusing on learning response from detection bounding boxes and classification predictions. Firstly, our method transferring category knowledge while equipping student model with the ability to retain localization knowledge during incremental learning. In addition, we further evaluate the qualities of all locations and provides valuable response by adaptive pseudo-label selection (APS) strategies. Finally, we elucidate that knowledge from different responses should be assigned with different importance during incremental distillation. Extensive experiments conducted on MS COCO demonstrate significant advantages of our method, which substantially narrow the performance gap towards full training.

show abstract

Section: Experiments and Discussionmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Response-based Distillation for Incremental Object Detection

Feng¹,

Wang²

2021

Preprint

View full text Add to dashboard Cite

show abstract

“…They additionally investigated the negative impact that having old class objects within the new class images has on the performance of the RPN and concluded that it was not that significant, which explains why Faster-RCNN networks generalize better than solutions with external proposals. Peng et al [91] presented the use of distillation not only on intermediate features, but also on the relations (distances) between features of different samples for anchor-free object detectors. Yang et al [70] proposed the preservation of channelwise, point-wise, and instance-wise correlations between some feature maps of the teacher and student networks in order to maintain the performance on the old classes while optimizing for the new ones.…”

Section: Knowledge Distillationmentioning

confidence: 99%

“…Knowledge Distillation: a strong baseline with caveats: It is easy to notice from Table 2 that most proposed strategies in the CIOD field use knowledge distillation as their primary mechanism to mitigate the effects of catastrophic forgetting. Comparing the results of the selected papers on the PASCAL VOC 2007 and MS COCO incremental benchmarks considering the metrics that assess the stability-plasticity of solutions, the differences between a recently proposed distillation technique such as Peng et al [91] and the first work of Shmelkov et al [7] are subtle. This either means that researchers might have been overfitting their solutions to the benchmarks or that simple logits and bounding box distillation are a strong baseline.…”

Section: Trends and Research Directionsmentioning

confidence: 99%

Continual Object Detection: A review of definitions, strategies, and challenges

Menezes¹,

Moura²,

Alves³

et al. 2022

Preprint

View full text Add to dashboard Cite

The field of Continual Learning investigates the ability to learn consecutive tasks without losing performance on those previously learned. Its focus has been mainly on incremental classification tasks. We believe that research in continual object detection deserves even more attention due to its vast range of applications in robotics and autonomous vehicles. This scenario is more complex than conventional classification given the occurrence of instances of classes that are unknown at the time, but can appear in subsequent tasks as a new class to be learned, resulting in missing annotations and conflicts with the background label. In this review, we analyze the current strategies proposed to tackle the problem of class-incremental object detection. Our main contributions are:(1) a short and systematic review of the methods that propose solutions to traditional incremental object detection scenarios; (2) A comprehensive evaluation of the existing approaches using a new metric to quantify the stability and plasticity of each technique in a standard way; (3) an overview of the current trends within continual object detection and a discussion of possible future research directions.

show abstract