Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
97 tokens/sec
GPT-4o
53 tokens/sec
Gemini 2.5 Pro Pro
43 tokens/sec
o3 Pro
4 tokens/sec
GPT-4.1 Pro
47 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

Enhancing Visual Domain Adaptation with Source Preparation (2306.10142v1)

Published 16 Jun 2023 in cs.CV, cs.AI, and cs.RO

Abstract: Robotic Perception in diverse domains such as low-light scenarios, where new modalities like thermal imaging and specialized night-vision sensors are increasingly employed, remains a challenge. Largely, this is due to the limited availability of labeled data. Existing Domain Adaptation (DA) techniques, while promising to leverage labels from existing well-lit RGB images, fail to consider the characteristics of the source domain itself. We holistically account for this factor by proposing Source Preparation (SP), a method to mitigate source domain biases. Our Almost Unsupervised Domain Adaptation (AUDA) framework, a label-efficient semi-supervised approach for robotic scenarios -- employs Source Preparation (SP), Unsupervised Domain Adaptation (UDA) and Supervised Alignment (SA) from limited labeled data. We introduce CityIntensified, a novel dataset comprising temporally aligned image pairs captured from a high-sensitivity camera and an intensifier camera for semantic segmentation and object detection in low-light settings. We demonstrate the effectiveness of our method in semantic segmentation, with experiments showing that SP enhances UDA across a range of visual domains, with improvements up to 40.64% in mIoU over baseline, while making target models more robust to real-world shifts within the target domain. We show that AUDA is a label-efficient framework for effective DA, significantly improving target domain performance with only tens of labeled samples from the target domain.

Definition Search Book Streamline Icon: https://streamlinehq.com
References (51)
  1. Randolph Blake. The visual system of the cat. Perception & Psychophysics, 1979.
  2. Integration of visual and infrared information in bimodal neurons in the rattlesnake optic tectum. Science, 1981.
  3. A theory of learning from different domains. Machine Learning, 2010.
  4. Unsupervised domain adaptation by backpropagation. In ICML, 2015.
  5. Unsupervised domain adaptation with residual transfer networks. NeurIPS, 2016.
  6. Maximum classifier discrepancy for unsupervised domain adaptation. In CVPR, 2018.
  7. Semi-supervised domain adaptation based on dual-level domain mixing for semantic segmentation. In CVPR, 2021.
  8. Semi-supervised dual-domain adaptation for semantic segmentation. ICPR, 2022.
  9. Ankit Singh. Clda: Contrastive learning for semi-supervised domain adaptation. NeurIPS, 2021.
  10. Semi-supervised domain adaptation via sample-to-sample self-distillation. In WACV, 2022.
  11. Few-shot adversarial domain adaptation. In NeurIPS, 2017.
  12. Domain-adaptive few-shot learning. In WACV, 2021.
  13. Pixel-by-pixel cross-domain alignment for few-shot semantic segmentation. In WACV, 2022.
  14. Few-shot structured domain adaptation for virtual-to-real scene parsing. In ICCV-Workshop, 2019.
  15. Refign: Align and refine for adaptation of semantic segmentation to adverse conditions. In WACV, 2023.
  16. Dannet: A one-stage domain adaptation network for unsupervised nighttime semantic segmentation. In CVPR, 2021.
  17. Map-guided curriculum domain adaptation and uncertainty-aware evaluation for semantic nighttime image segmentation. TPAMI, 2020.
  18. Iterative loop method combining active and semi-supervised learning for domain adaptive semantic segmentation. arXiv preprint arXiv:2301.13361, 2023.
  19. Mic: Masked image consistency for context-enhanced domain adaptation. In CVPR, 2023.
  20. A survey of unsupervised deep domain adaptation. TIST, 2020.
  21. Fcns in the wild: Pixel-level adversarial and constraint-based adaptation. In arXiv preprint arXiv:1612.02649, 2016.
  22. CyCADA: Cycle-consistent adversarial domain adaptation. In ICML, 2018.
  23. Learning to adapt structured output space for semantic segmentation. In CVPR, 2018.
  24. Advent: Adversarial entropy minimization for domain adaptation in semantic segmentation. In CVPR, 2019.
  25. Domain-adversarial training of neural networks. JMLR, 2016.
  26. Bidirectional learning for domain adaptation of semantic segmentation. In CVPR, 2019.
  27. Learning texture invariant representation for domain adaptation of semantic segmentation. In CVPR, 2020.
  28. Unsupervised domain adaptation for semantic segmentation via class-balanced self-training. In ECCV, 2018.
  29. Classes matter: A fine-grained adversarial approach to cross-domain semantic segmentation. In ECCV, 2020.
  30. Dong-Hyun Lee. Pseudo-label : The simple and efficient semi-supervised learning method for deep neural networks. ICML Workshop, 2013.
  31. Generalizing from several related classification tasks to a new unlabeled sample. In NeurIPS, 2011.
  32. Style normalization and restitution for domain generalization and adaptation. IEEE Transactions on Multimedia, 2021.
  33. Domain-specific bias filtering for single labeled domain generalization. IJCV, 2023.
  34. Dengxin Dai and Luc Van Gool. Dark model adaptation: Semantic image segmentation from daytime to nighttime. In ITSC, 2018.
  35. Bdd100k: A diverse driving dataset for heterogeneous multitask learning. In CVPR, 2020.
  36. Mfnet: Towards real-time semantic segmentation for autonomous vehicles with multi-spectral scenes. In IROS, 2017.
  37. Domain generalization with mixstyle. In ICLR, 2021.
  38. Segformer: Simple and efficient design for semantic segmentation with transformers. In NeurIPS, 2021.
  39. Arbitrary style transfer in real-time with adaptive instance normalization. In ICCV, 2017.
  40. Image style transfer using convolutional neural networks. In CVPR, 2016.
  41. Demystifying neural style transfer. In IJCAI, 2017.
  42. Instaformer: Instance-aware image-to-image translation with transformer. In CVPR, 2022.
  43. mixup: Beyond empirical risk minimization. ICLR, 2017.
  44. Regmixup: Mixup as a regularizer can surprisingly improve accuracy and out distribution robustness. In NeurIPS, 2023.
  45. A survey on image data augmentation for deep learning. Journal of Big Data, 2019.
  46. Segment anything. arXiv:2304.02643, 2023.
  47. The cityscapes dataset for semantic urban scene understanding. In CVPR, 2016.
  48. imgaug. https://github.com/aleju/imgaug, 2020.
  49. Universal semi-supervised semantic segmentation. In ICCV, 2019.
  50. Dilated residual networks. In CVPR, 2017.
  51. NVIDIA. Nvidia a100 gpu. https://www.nvidia.com/en-us/data-center/a100/.
User Edit Pencil Streamline Icon: https://streamlinehq.com
Authors (4)
  1. Anirudha Ramesh (4 papers)
  2. Anurag Ghosh (11 papers)
  3. Christoph Mertz (10 papers)
  4. Jeff Schneider (99 papers)
Citations (1)
View on Semantic Scholar

Summary

We haven't generated a summary for this paper yet.

Ai Sparkles Streamline Icon: https://streamlinehq.com Summarize Now