Domain Adaptive Knowledge Distillation for Driving Scene Semantic Segmentation
Divya Kothandaraman,Athira M. Nambiar,Anurag Mittal +2 more
- pp 134-143
Reads0
Chats0
TLDR
In this article, a multi-level distillation strategy is proposed to effectively distil knowledge at different levels, and a novel cross entropy loss is introduced to leverage pseudo labels from the teacher.Abstract:
Practical autonomous driving systems face two crucial challenges: memory constraints and domain gap issues. In this paper, we present a novel approach to learn domain adaptive knowledge in models with limited memory, thus bestowing the model with the ability to deal with these issues in a comprehensive manner. We term this as “Domain Adaptive Knowledge Distillation ” and address the same in the context of unsupervised domain-adaptive semantic segmentation by proposing a multi-level distillation strategy to effectively distil knowledge at different levels. Further, we introduce a novel cross entropy loss that leverages pseudo labels from the teacher. These pseudo teacher labels play a multifaceted role towards: (i) knowledge distillation from the teacher network to the student network & (ii) serving as a proxy for the ground truth for target domain images, where the problem is completely unsupervised. We introduce four paradigms for distilling domain adaptive knowledge and carry out extensive experiments and ablation studies on real-to-real as well as synthetic-to-real scenarios. Our experiments demonstrate the profound success of our proposed method.read more
Citations
More filters
Journal ArticleDOI
Confidence-and-Refinement Adaptation Model for Cross-Domain Semantic Segmentation
TL;DR: A novel multi-level UDA model named Confidence-and-Refinement Adaptation Model (CRAM), which contains a confidence-aware entropy alignment (CEA) module and a style feature alignment (SFA) module, which achieves comparable performance with the existing state-of-the-art works with advantages in simplicity and convergence speed.
Journal ArticleDOI
Robust Semantic Segmentation With Multi-Teacher Knowledge Distillation
TL;DR: In this paper, a multi-teacher knowledge distillation (KD) framework was proposed to address time-consuming annotation task in semantic segmentation, through which one teacher trained on a single dataset could be leveraged for annotating unlabeled data.
Proceedings ArticleDOI
Multi-Domain Incremental Learning for Semantic Segmentation
TL;DR: In this paper , the authors propose a dynamic architecture that assigns universally shared, domain-invariant parameters to capture homogeneous semantic features present in all domains, while dedicated domain-specific parameters learn the statistics of each domain.
Posted Content
Multi-Domain Incremental Learning for Semantic Segmentation
Prachi Garg,Rohit Saluja,Vineeth N Balasubramanian,Chetan Arora,Anbumani Subramanian,C. V. Jawahar +5 more
TL;DR: In this paper, the authors propose a dynamic architecture that assigns universally shared, domain-invariant parameters to capture homogeneous semantic features present in all domains, while dedicated domain-specific parameters learn the statistics of each domain.
Book ChapterDOI
Neural Network Compression Through Shunt Connections and Knowledge Distillation for Semantic Segmentation Problems
TL;DR: Shunt connections are used in this article for MobileNet compression and segmentation tasks on the Cityscapes dataset, on which they achieve compression by 28% while observing a 3.52 drop in mIoU.
References
More filters
Proceedings ArticleDOI
Structured Knowledge Distillation for Semantic Segmentation
TL;DR: Zhang et al. as mentioned in this paper investigated the issue of knowledge distillation for training compact semantic segmentation networks by making use of cumbersome networks and proposed to distill the structured knowledge from cumbersome networks into compact networks.
Proceedings ArticleDOI
On the Efficacy of Knowledge Distillation
Jang Hyun Cho,Bharath Hariharan +1 more
TL;DR: It is found crucially that larger models do not often make better teachers, and that small students are unable to mimic large teachers.
Proceedings ArticleDOI
CrDoCo: Pixel-Level Domain Transfer With Cross-Domain Consistency
TL;DR: A novel pixel-wise adversarial domain adaptation algorithm that leverages image-to-image translation methods for data augmentation and introduces a cross-domain consistency loss that enforces the adapted model to produce consistent predictions.
Proceedings ArticleDOI
MARS: Motion-Augmented RGB Stream for Action Recognition
TL;DR: This paper introduces two learning approaches to train a standard 3D CNN, operating on RGB frames, that mimics the motion stream, and as a result avoids flow computation at test time, and denotes the stream trained using this combined loss as Motion-Augmented RGB Stream (MARS).
Proceedings Article
Mutual Mean-Teaching: Pseudo Label Refinery for Unsupervised Domain Adaptation on Person Re-identification
TL;DR: In this article, the authors proposed an unsupervised framework, Mutual Mean-Teaching (MMT), to learn better features from the target domain via off-line refined hard pseudo labels and online refined soft pseudo labels in an alternative training manner.