Re-examining Distillation For Continual Object Detection

• URL: http://arxiv.org/abs/2204.01407v1
• Date: Mon, 4 Apr 2022 11:50:54 GMT
• Title: Re-examining Distillation For Continual Object Detection
• Authors: Eli Verwimp, Kuo Yang, Sarah Parisot, Hong Lanqing, Steven McDonagh, Eduardo P\'erez-Pellitero, Matthias De Lange and Tinne Tuytelaars
• Abstract summary: We conduct a thorough analysis of why object detection models forget catastrophically. We focus on distillation-based approaches in two-stage networks. We show that this works well for the region proposal network, but that wrong, yet overly confident teacher predictions prevent student models from effective learning of the classification head.
• Score: 33.95470797472666
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Training models continually to detect and classify objects, from new classes and new domains, remains an open problem. In this work, we conduct a thorough analysis of why and how object detection models forget catastrophically. We focus on distillation-based approaches in two-stage networks; the most-common strategy employed in contemporary continual object detection work.Distillation aims to transfer the knowledge of a model trained on previous tasks -- the teacher -- to a new model -- the student -- while it learns the new task. We show that this works well for the region proposal network, but that wrong, yet overly confident teacher predictions prevent student models from effective learning of the classification head. Our analysis provides a foundation that allows us to propose improvements for existing techniques by detecting incorrect teacher predictions, based on current ground-truth labels, and by employing an adaptive Huber loss as opposed to the mean squared error for the distillation loss in the classification heads. We evidence that our strategy works not only in a class incremental setting, but also in domain incremental settings, which constitute a realistic context, likely to be the setting of representative real-world problems.

Related papers

• Decoupling the Class Label and the Target Concept in Machine Unlearning [81.69857244976123]
  Machine unlearning aims to adjust a trained model to approximate a retrained one that excludes a portion of training data. Previous studies showed that class-wise unlearning is successful in forgetting the knowledge of a target class. We propose a general framework, namely, TARget-aware Forgetting (TARF)
  arXiv  Detail & Related papers  (2024-06-12T14:53:30Z)
• Task Integration Distillation for Object Detectors [2.974025533366946]
  We propose a knowledge distillation method that addresses both the classification and regression tasks. We evaluate the importance of features based on the output of the detector's two sub-tasks. This method effectively prevents the issue of biased predictions about the model's learning reality.
  arXiv  Detail & Related papers  (2024-04-02T07:08:15Z)
• Activate and Reject: Towards Safe Domain Generalization under Category Shift [71.95548187205736]
  We study a practical problem of Domain Generalization under Category Shift (DGCS) It aims to simultaneously detect unknown-class samples and classify known-class samples in the target domains. Compared to prior DG works, we face two new challenges: 1) how to learn the concept of unknown'' during training with only source known-class samples, and 2) how to adapt the source-trained model to unseen environments.
  arXiv  Detail & Related papers  (2023-10-07T07:53:12Z)
• EvCenterNet: Uncertainty Estimation for Object Detection using Evidential Learning [26.535329379980094]
  EvCenterNet is a novel uncertainty-aware 2D object detection framework. We employ evidential learning to estimate both classification and regression uncertainties. We train our model on the KITTI dataset and evaluate it on challenging out-of-distribution datasets.
  arXiv  Detail & Related papers  (2023-03-06T11:07:11Z)
• TIDo: Source-free Task Incremental Learning in Non-stationary Environments [0.0]
  Updating a model-based agent to learn new target tasks requires us to store past training data. Few-shot task incremental learning methods overcome the limitation of labeled target datasets. We propose a one-shot task incremental learning approach that can adapt to non-stationary source and target tasks.
  arXiv  Detail & Related papers  (2023-01-28T02:19:45Z)
• Bridging Non Co-occurrence with Unlabeled In-the-wild Data for Incremental Object Detection [56.22467011292147]
  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. We propose the use of unlabeled in-the-wild data to bridge the non-occurrence caused by the missing base classes during the training of additional novel classes.
  arXiv  Detail & Related papers  (2021-10-28T10:57:25Z)
• Incrementally Zero-Shot Detection by an Extreme Value Analyzer [0.0]
  This paper introduces a novel strategy for both zero-shot learning and class-incremental learning in real-world object detection. We propose a novel extreme value analyzer to detect objects from old seen, new seen, and unseen classes, simultaneously. Experiments demonstrate the efficacy of our model in detecting objects from both the seen and unseen classes, outperforming the alternative models on Pascal VOC and MSCOCO datasets.
  arXiv  Detail & Related papers  (2021-03-23T15:06:30Z)
• Open-set Short Utterance Forensic Speaker Verification using Teacher-Student Network with Explicit Inductive Bias [59.788358876316295]
  We propose a pipeline solution to improve speaker verification on a small actual forensic field dataset. By leveraging large-scale out-of-domain datasets, a knowledge distillation based objective function is proposed for teacher-student learning. We show that the proposed objective function can efficiently improve the performance of teacher-student learning on short utterances.
  arXiv  Detail & Related papers  (2020-09-21T00:58:40Z)
• Lifelong Object Detection [28.608982224098565]
  We leverage the fact that new training classes arrive in a sequential manner and incrementally refine the model. We consider the representative object detector, Faster R-CNN, for both accurate and efficient prediction.
  arXiv  Detail & Related papers  (2020-09-02T15:08:51Z)
• Distilling Object Detectors with Task Adaptive Regularization [97.52935611385179]
  Current state-of-the-art object detectors are at the expense of high computational costs and are hard to deploy to low-end devices. Knowledge distillation, which aims at training a smaller student network by transferring knowledge from a larger teacher model, is one of the promising solutions for model miniaturization.
  arXiv  Detail & Related papers  (2020-06-23T15:58:22Z)
• Incremental Object Detection via Meta-Learning [77.55310507917012]
  We propose a meta-learning approach that learns to reshape model gradients, such that information across incremental tasks is optimally shared. In comparison to existing meta-learning methods, our approach is task-agnostic, allows incremental addition of new-classes and scales to high-capacity models for object detection.
  arXiv  Detail & Related papers  (2020-03-17T13:40:00Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.