Unsupervised Domain Adaptive 3D Detection with Multi-Level Consistency

Luo, Zhipeng; Cai, Zhongang; Zhou, Changqing; Zhang, Gongjie; Zhao, Haiyu; Yi, Shuai; Lu, Shijian; Li, Hongsheng; Zhang, Shanghang; Liu, Ziwei

doi:10.48550/arxiv.2107.11355

Cited by 4 publications

(6 citation statements)

References 36 publications

(68 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As a result, when the detector is adapted to a sparse domain, it can not generate enough highquality pseudo labels to provide sufficient knowledge in the selftraining stage. TABLE 4 Unsupervised adaptation results of SF-UDA 3D [23], Dreaming [24], MLC-Net [26] and our ST3D++. We report AP 3D of car at IoU 0.7 and 0.5 on nuScenes → KITTI.…”

Section: Main Results and Analysismentioning

confidence: 99%

“…We report AP 3D of car at IoU 0.7 and 0.5 on nuScenes → KITTI. To further demonstrate the advancement of our ST3D++, we compare it with SF-UDA 3D [23], Dreaming [24] and MLC-Net [26] on nuScenes → KITTI (i.e. the only common adaptation 2.…”

Section: Main Results and Analysismentioning

confidence: 99%

“…(vi) We conduct extensive experiments on four datasets for three categories. The proposed ST3D++ outperforms ST3D [25] over all categories on all adaptation settings as well as other existing UDA methods, SF-UDA 3D [23], Dreaming [24] and MLC-Net [26] by a large margin. (vii) We explore to leverage temporal information to further improve ST3D++ through fused sequential point frames, which also mitigates the point density gaps across domains.…”

Section: Introductionmentioning

confidence: 84%

“…SRDAN [64] employed scale-aware and rangeaware feature adversarial alignment manners to match the distribution between two domains, which might suffer from stability and convergence issues. MLC-Net [26] employed the meanteacher paradigm to address the geometric mismatch between source and target domains. In contrast, the proposed ST3D++ tailors a self-denoising framework to simultaneously close content and point distribution gaps across domains and achieve superior performance on all three categories of four adaptation tasks with neither prior target object statistics nor extra computation.…”

Section: Related Workmentioning

confidence: 99%

See 3 more Smart Citations

ST3D++: Denoised Self-training for Unsupervised Domain Adaptation on 3D Object Detection

Yang¹,

Shi²,

Wang³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

In this paper, we present a self-training method, named ST3D++, with a holistic pseudo label denoising pipeline for unsupervised domain adaptation on 3D object detection. ST3D++ aims at reducing noise in pseudo label generation as well as alleviating the negative impacts of noisy pseudo labels on model training. First, ST3D++ pre-trains the 3D object detector on the labeled source domain with random object scaling (ROS) which is designed to reduce target domain pseudo label noise arising from object scale bias of the source domain. Then, the detector is progressively improved through alternating between generating pseudo labels and training the object detector with pseudo-labeled target domain data. Here, we equip the pseudo label generation process with a hybrid quality-aware triplet memory to improve the quality and stability of generated pseudo labels. Meanwhile, in the model training stage, we propose a source data assisted training strategy and a curriculum data augmentation policy to effectively rectify noisy gradient directions and avoid model over-fitting to noisy pseudo labeled data. These specific designs enable the detector to be trained on meticulously refined pseudo labeled target data with denoised training signals, and thus effectively facilitate adapting an object detector to a target domain without requiring annotations. Finally, our method is assessed on four 3D benchmark datasets (i.e. Waymo, KITTI, Lyft, and nuScenes) for three common categories (i.e. car, pedestrian and bicycle). ST3D++ achieves state-of-the-art performance on all evaluated settings, outperforming the corresponding baseline by a large margin (e.g. 9.6% ∼ 38.16% on Waymo → KITTI in terms of AP 3D ), and even surpasses the fully supervised oracle results on the KITTI 3D object detection benchmark with target prior. Code will be available.

show abstract

Section: Main Results and Analysismentioning

confidence: 99%

Section: Main Results and Analysismentioning

confidence: 99%

Section: Introductionmentioning

confidence: 84%

Section: Related Workmentioning

confidence: 99%

See 2 more Smart Citations

ST3D++: Denoised Self-training for Unsupervised Domain Adaptation on 3D Object Detection

Yang¹,

Shi²,

Wang³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…The second category is image translation-based which adapts image appearance to mitigate domain gaps [7,14,241,295]. The third category is self-training-based which predicts pseudo labels or minimizes entropy to guide iterative learning from target samples [13,141,183,184,242,296].…”

Section: Discussionmentioning

confidence: 99%

Transductive transfer learning for visual recognition

Huang¹

View full text Add to dashboard Cite

has given invaluable advice on almost all aspects of my Ph.D. study. I would also express great gratitude towards all the lab-mates in MICL, CSL and SCALE Labs and all the members in Prof Lu's research group for the enthusiastic discussions and priceless support in the last four years. Dayan, Aoran and Pengdeng brought me a lot of precious memory in my Ph.D. journey. Thank Jingyi, Kaiwen, Zichen, Yun and Xueying for their hardworking while we collaborated on projects and researches. I enjoy the good works we have done and the team-working time.Thanks to all of my thesis committee members, Prof. Lin Guosheng, Prof. Liu Ziwei and Prof. Chen Change Loy for the time, kind suggestions and feedback for my qualification examination and thesis. You have all been very patient and kind throughout the whole process, and your encouragement and suggestions have helped refine my work greatly.Last but most importantly, I wish to express the deepest gratitude to my family for their constant and unconditional love and support. Thank you for being always there for me, for listening, for enlightening and for giving me the strength and courage to overcome difficulties and pursue greatness.xi

show abstract

Model Adaptation: Historical Contrastive Learning for Unsupervised Domain Adaptation without Source Data

Huang¹,

Guan²,

Xiao³

et al. 2021

Preprint

View full text Add to dashboard Cite

Unsupervised domain adaptation aims to align a labeled source domain and an unlabeled target domain, but it requires to access the source data which often raises concerns in data privacy, data portability and data transmission efficiency. We study unsupervised model adaptation (UMA), or called Unsupervised Domain Adaptation without Source Data, an alternative setting that aims to adapt source-trained models towards target distributions without accessing source data. To this end, we design an innovative historical contrastive learning (HCL) technique that exploits historical source hypothesis to make up for the absence of source data in UMA. HCL addresses the UMA challenge from two perspectives. First, it introduces historical contrastive instance discrimination (HCID) that learns from target samples by contrasting their embeddings which are generated by the currently adapted model and the historical models. With the historical models, HCID encourages UMA to learn instance-discriminative target representations while preserving the source hypothesis. Second, it introduces historical contrastive category discrimination (HCCD) that pseudo-labels target samples to learn category-discriminative target representations. Specifically, HCCD re-weights pseudo labels according to their prediction consistency across the current and historical models. Extensive experiments show that HCL outperforms and state-of-the-art methods consistently across a variety of visual tasks and setups.

show abstract

Unsupervised Domain Adaptive 3D Detection with Multi-Level Consistency

Cited by 4 publications

References 36 publications

ST3D++: Denoised Self-training for Unsupervised Domain Adaptation on 3D Object Detection

ST3D++: Denoised Self-training for Unsupervised Domain Adaptation on 3D Object Detection

Transductive transfer learning for visual recognition

Model Adaptation: Historical Contrastive Learning for Unsupervised Domain Adaptation without Source Data

Contact Info

Product

Resources

About