Korea Advanced Institute of Science and Technology · Computer Science
Professor Thang Vu's research lab specializes in advancing 3D vision and object detection, with a strong focus on improving instance segmentation and region proposal networks. The lab develops novel deep learning architectures—such as SoftGroup and Cascade RPN—that address fundamental limitations in semantic prediction and anchor design through soft grouping, multi-stage refinement, and improved feature alignment. By emphasizing uncertainty mitigation, training-inference distribution consistency, and contextual feature learning, the lab aims to enhance both accuracy and scalability in 3D perception systems. Their work bridges the gap between theoretical robustness and practical deployment in real-world vision applications.
Figures are computed from collected data and may differ slightly.
Existing state-of-the-art 3D instance segmentation methods perform semantic segmentation followed by grouping. The hard predictions are made when performing semantic segmentation such that each point is associated with a single class. However, the errors stemming from hard decision propagate into grouping that results in (1) low overlaps between the predicted instance with the ground truth and (2) substantial false positives. To address the aforementioned problems, this paper proposes a 3D insta
Cascaded architectures have brought significant performance improvement in object detection and instance segmentation. However, there are lingering issues regarding the disparity in the Intersection-over-Union (IoU) distribution of the samples between training and inference. This disparity can potentially exacerbate detection accuracy. This paper proposes an architecture referred to as Sample Consistency Network (SCNet) to ensure that the IoU distribution of the samples at training time is close
This paper considers an architecture referred to as Cascade Region Proposal\nNetwork (Cascade RPN) for improving the region-proposal quality and detection\nperformance by \\textit{systematically} addressing the limitation of the\nconventional RPN that \\textit{heuristically defines} the anchors and\n\\textit{aligns} the features to the anchors. First, instead of using multiple\nanchors with predefined scales and aspect ratios, Cascade RPN relies on a\n\\textit{single anchor} per location and per
This paper considers an architecture referred to as Cascade Region Proposal Network (Cascade RPN) for improving the region-proposal quality and detection performance by \textit{systematically} addressing the limitation of the conventional RPN that \textit{heuristically defines} the anchors and \textit{aligns} the features to the anchors. First, instead of using multiple anchors with predefined scales and aspect ratios, Cascade RPN relies on a \textit{single anchor} per location and performs mult
This paper considers a network referred to as SoftGroup for accurate and scalable 3D instance segmentation. Existing state-of-the-art methods produce hard semantic predictions followed by grouping instance segmentation results. Unfortunately, errors stemming from hard decisions propagate into the grouping, resulting in poor overlap between predicted instances and ground truth and substantial false positives. To address the abovementioned problems, SoftGroup allows each point to be associated wit
Cascaded architectures have brought significant performance improvement in object detection and instance segmentation. However, there are lingering issues regarding the disparity in the Intersection-over-Union (IoU) distribution of the samples between training and inference. This disparity can potentially exacerbate detection accuracy. This paper proposes an architecture referred to as Sample Consistency Network (SCNet) to ensure that the IoU distribution of the samples at training time is close
A bounding box commonly serves as the proxy for 2D object detection. However, extending this practice to 3D detection raises sensitivity to localization error. This problem is acute on flat objects since small localization error may lead to low overlaps between the prediction and ground truth. To address this problem, this paper proposes Sphere Region Proposal Network (SphereRPN) which detects objects by learning spheres as opposed to bounding boxes. We demonstrate that spherical proposals are m
Open papers in the app to read, cite, and organize with AI.
