Continual detection transformer for incremental object detection
Incremental object detection (IOD) aims to train an object detector in phases, each with annotations for new object categories. As other incremental settings, IOD is subject to catastrophic forgetting, which is often addressed by techniques such as knowledge distillation (KD) and exemplar replay (ER...
| Autores principales: | , , , |
|---|---|
| Formato: | Conference item |
| Lenguaje: | English |
| Publicado: |
IEEE
2023
|
| _version_ | 1826311455371165696 |
|---|---|
| author | Liu, Y Schiele, B Vedaldi, A Rupprecht, C |
| author_facet | Liu, Y Schiele, B Vedaldi, A Rupprecht, C |
| author_sort | Liu, Y |
| collection | OXFORD |
| description | Incremental object detection (IOD) aims to train an object detector in phases, each with annotations for new object categories. As other incremental settings, IOD is subject to catastrophic forgetting, which is often addressed by techniques such as knowledge distillation (KD) and exemplar replay (ER). However, KD and ER do not work well if applied directly to state-of-the-art transformer-based object detectors such as Deformable DETR and UPDETR. In this paper, we solve these issues by proposing a ContinuaL DEtection TRansformer (CL-DETR), a new method for transformer-based IOD which enables effective usage of KD and ER in this context. First, we introduce a Detector Knowledge Distillation (DKD) loss, focusing on the most informative and reliable predictions from old versions of the model, ignoring redundant background predictions, and ensuring compatibility with the available groundtruth labels. We also improve ER by proposing a calibration strategy to preserve the label distribution of the training set, therefore better matching training and testing statistics. We conduct extensive experiments on COCO 2017 and demonstrate that CL-DETR achieves state-of-the-art results in the IOD setting.
|
| first_indexed | 2024-03-07T08:10:06Z |
| format | Conference item |
| id | oxford-uuid:c4e0ac4b-210d-4a81-b75d-4d0e5526ab8a |
| institution | University of Oxford |
| language | English |
| last_indexed | 2024-03-07T08:10:06Z |
| publishDate | 2023 |
| publisher | IEEE |
| record_format | dspace |
| spelling | oxford-uuid:c4e0ac4b-210d-4a81-b75d-4d0e5526ab8a2023-11-17T11:09:52ZContinual detection transformer for incremental object detection Conference itemhttp://purl.org/coar/resource_type/c_5794uuid:c4e0ac4b-210d-4a81-b75d-4d0e5526ab8aEnglishSymplectic ElementsIEEE2023Liu, YSchiele, BVedaldi, ARupprecht, CIncremental object detection (IOD) aims to train an object detector in phases, each with annotations for new object categories. As other incremental settings, IOD is subject to catastrophic forgetting, which is often addressed by techniques such as knowledge distillation (KD) and exemplar replay (ER). However, KD and ER do not work well if applied directly to state-of-the-art transformer-based object detectors such as Deformable DETR and UPDETR. In this paper, we solve these issues by proposing a ContinuaL DEtection TRansformer (CL-DETR), a new method for transformer-based IOD which enables effective usage of KD and ER in this context. First, we introduce a Detector Knowledge Distillation (DKD) loss, focusing on the most informative and reliable predictions from old versions of the model, ignoring redundant background predictions, and ensuring compatibility with the available groundtruth labels. We also improve ER by proposing a calibration strategy to preserve the label distribution of the training set, therefore better matching training and testing statistics. We conduct extensive experiments on COCO 2017 and demonstrate that CL-DETR achieves state-of-the-art results in the IOD setting. |
| spellingShingle | Liu, Y Schiele, B Vedaldi, A Rupprecht, C Continual detection transformer for incremental object detection |
| title | Continual detection transformer for incremental object detection |
| title_full | Continual detection transformer for incremental object detection |
| title_fullStr | Continual detection transformer for incremental object detection |
| title_full_unstemmed | Continual detection transformer for incremental object detection |
| title_short | Continual detection transformer for incremental object detection |
| title_sort | continual detection transformer for incremental object detection |
| work_keys_str_mv | AT liuy continualdetectiontransformerforincrementalobjectdetection AT schieleb continualdetectiontransformerforincrementalobjectdetection AT vedaldia continualdetectiontransformerforincrementalobjectdetection AT rupprechtc continualdetectiontransformerforincrementalobjectdetection |