Lifelong Object Detection

Zhou, Wang; Chang, Shiyu; Sosa, Norma; Hamann, Hendrik; Cox, David

doi:10.48550/arxiv.2009.01129

Cited by 11 publications

(18 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…[27] on the novel classes when trained without co-occurrence (Ours "w category": 62.2%; Ours "w/o category": 62.3% vs. [27]: 49.1%). In Tables 3 and 4, we also show that our method outperforms [27] and [40] under the "w co − occur" setting. These results show the effectiveness of our proposed approach when a group of novel classes are added.…”

Section: Dual-teacher Distillationmentioning

confidence: 64%

“…We can see that our method shows on par performance with [27] under [27]: 47.3%). We also outperform [40] significantly under the "w co − occur" setting.…”

Section: Comparison Of Methodsmentioning

confidence: 88%

“…Interestingly, we can also run our dual-teacher distillation framework on the "w co − occur" data. It can be seen that we significantly outperform [27] and [40]. [27] on the novel classes when trained without co-occurrence (Ours "w category": 62.2%; Ours "w/o category": 62.3% vs. [27]: 49.1%).…”

Section: Dual-teacher Distillationmentioning

confidence: 88%

“…Zhou et. al [40] proposes an incremental version of Faster R-CNN that distills selected anchors and proposals with the Pseudo-Positive-Aware Sampling strategy. However, it fails to improve the benchmark [27] on the standard evaluation criteria.…”

Section: Class-incremental Object Detectionmentioning

confidence: 99%

“…Recently, several incremental learning methods [40,27,4,10,20,31] are proposed to overcome catastrophic forgetting in the object detection task. Despite the impressive performance, these methods rely on the co-occurrence of unlabeled base classes in the training data of the novel classes.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Bridging Non Co-occurrence with Unlabeled In-the-wild Data for Incremental Object Detection

Dong¹,

Zhang²,

Ding³

et al. 2021

Preprint

View full text Add to dashboard Cite

Deep networks have shown remarkable results in the task of object detection. However, their performance suffers critical drops when they are subsequently trained on novel classes without any sample from the base classes originally used to train the model. This phenomenon is known as catastrophic forgetting. Recently, several incremental learning methods are proposed to mitigate catastrophic forgetting for object detection. Despite the effectiveness, these methods require co-occurrence of the unlabeled base classes in the training data of the novel classes. This requirement is impractical in many real-world settings since the base classes do not necessarily co-occur with the novel classes. In view of this limitation, we consider a more practical setting of complete absence of cooccurrence of the base and novel classes for the object detection task. We propose the use of unlabeled in-the-wild data to bridge the non co-occurrence caused by the missing base classes during the training of additional novel classes. To this end, we introduce a blind sampling strategy based on the responses of the baseclass model and pre-trained novel-class model to select a smaller relevant dataset from the large in-the-wild dataset for incremental learning. We then design a dual-teacher distillation framework to transfer the knowledge distilled from the base-and novel-class teacher models to the student model using the sampled inthe-wild data. Experimental results on the PASCAL VOC and MS COCO datasets show that our proposed method significantly outperforms other state-of-the-art class-incremental object detection methods when there is no co-occurrence between the base and novel classes during training. Our source code is available at https://github.com/dongnana777/Bridging-Non-Co-occurrence.

show abstract

Section: Dual-teacher Distillationmentioning

confidence: 64%

“…We can see that our method shows on par performance with [27] under [27]: 47.3%). We also outperform [40] significantly under the "w co − occur" setting.…”

Section: Comparison Of Methodsmentioning

confidence: 88%

Section: Dual-teacher Distillationmentioning

confidence: 88%

Section: Class-incremental Object Detectionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Bridging Non Co-occurrence with Unlabeled In-the-wild Data for Incremental Object Detection

Dong¹,

Zhang²,

Ding³

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

Continual semi-supervised learning through contrastive interpolation consistency

Boschini

Buzzega

Bonicelli

et al. 2022

Pattern Recognition Letters

View full text Add to dashboard Cite

Comparative Study of Natural Replay and Experience Replay in Online Object Detection

Wagner,

Pellerin,

Huet

2023

2023 IEEE/CVF International Conference on Computer Vision Workshops (ICCVW)

View full text Add to dashboard Cite

Online Object Detection (OOD) algorithms play a crucial role in dynamic and real-world computer vision applications. In these scenarios, models are trained on a data stream where old class samples are revisited, a phenomenon known as Natural Replay (NR). During training, NR occurs unevenly across object categories, leading to evaluation metrics biased towards the most frequently revisited classes. Existing benchmarks lack proper quantification of NR and depict short-term training scenarios on a single domain. As a result, evaluating generalization capabilities and forgetting rates of models become challenging in OOD. In this paper, we address the challenges surrounding the evaluation of OOD models by proposing two key contributions. Firstly, we define a metric to quantify NR in an OOD scenario and show how NR is related to class specific forgetting. Secondly, we introduce a novel benchmark, EgOAK, which introduces a long-term training scenario that involves frequent domain shifts. It allows the evaluation of models' generalization capabilities and forgetting of knowledge on past domains. Our results in this OOD setting reveal that Experience Replay, a memory-based method, is particularly effective for better generalization to new domains and for preserving past knowledge. Leveraging replay from memory helps to address the low natural replay rate for rarely revisited classes, resulting in improved adaptability and reliability of models in dynamic environments.

show abstract

Lifelong Object Detection

Cited by 11 publications

References 24 publications

Bridging Non Co-occurrence with Unlabeled In-the-wild Data for Incremental Object Detection

Bridging Non Co-occurrence with Unlabeled In-the-wild Data for Incremental Object Detection

Continual semi-supervised learning through contrastive interpolation consistency

Comparative Study of Natural Replay and Experience Replay in Online Object Detection

Contact Info

Product

Resources

About