no code implementations • 23 Dec 2013 • Prateek Singhal, Aditya Deshpande, N. Dinesh Reddy, K. Madhava Krishna
to perform better classification and merging .
no code implementations • 24 Apr 2015 • N. Dinesh Reddy, Prateek Singhal, K. Madhava Krishna
We pro- pose an algorithm that jointly infers the semantic class and motion labels of an object.
no code implementations • 27 Apr 2015 • N. Dinesh Reddy, Prateek Singhal, Visesh Chari, K. Madhava Krishna
We show results on the challenging KITTI urban dataset for accuracy of motion segmentation and reconstruction of the trajectory and shape of moving objects relative to ground truth.
no code implementations • 2 Aug 2016 • N. Dinesh Reddy, Iman Abbasnejad, Sheetal Reddy, Amit Kumar Mondal, Vindhya Devalla
Real time outdoor navigation in highly dynamic environments is an crucial problem.
no code implementations • 18 Apr 2017 • Nazrul Haque, N. Dinesh Reddy, K. Madhava Krishna
This paper proposes an approach to fuse semantic features and motion clues using CNNs, to address the problem of monocular semantic motion segmentation.
1 code implementation • CVPR 2018 • N. Dinesh Reddy, Minh Vo, Srinivasa G. Narasimhan
In this work, we develop a framework to fuse both the single-view feature tracks and multi-view detected part locations to significantly improve the detection, localization and reconstruction of moving vehicles, even in the presence of strong occlusions.
1 code implementation • CVPR 2019 • N. Dinesh Reddy, Minh Vo, Srinivasa G. Narasimhan
Central to this work is a trifocal tensor loss that provides indirect self-supervision for occluded keypoint locations that are visible in other views of the object.
Ranked #1 on Vehicle Pose Estimation on CarFusion
3D Car Instance Understanding 3D Object Reconstruction From A Single Image +2
no code implementations • CVPR 2021 • N. Dinesh Reddy, Laurent Guigues, Leonid Pischulini, Jayan Eledath, Srinivasa Narasimhan
At the core of our approach is a novel spatio-temporal formulation that operates in a common voxelized feature space aggregated from single- or multiple camera views.
Ranked #1 on 3D Human Pose Estimation on Panoptic (using extra training data)
1 code implementation • IEEE Intelligent Vehicles Symposium 2021 • Fangyu Li, N. Dinesh Reddy, Xudong Chen and Srinivasa G. Narasimhan
Reconstructing 4D vehicular activity (3D space and time) from cameras is useful for autonomous vehicles, commuters and local authorities to plan for smarter and safer cities.
1 code implementation • CVPR 2022 • N. Dinesh Reddy, Robert Tamburo, Srinivasa G. Narasimhan
Labeled real data of occlusions is scarce (even in large datasets) and synthetic data leaves a domain gap, making it hard to explicitly model and learn occlusions.
Ranked #1 on Amodal Instance Segmentation on WALT
no code implementations • CVPR 2023 • Anurag Ghosh, N. Dinesh Reddy, Christoph Mertz, Srinivasa G. Narasimhan
For autonomous navigation, using the same detector and scale, our approach improves detection rate by +4. 1 $AP_{S}$ or +39% and in real-time performance by +5. 3 $sAP_{S}$ or +63% for small objects over state-of-the-art (SOTA).
no code implementations • 27 Mar 2024 • Khiem Vuong, N. Dinesh Reddy, Robert Tamburo, Srinivasa G. Narasimhan
Current methods for 2D and 3D object understanding struggle with severe occlusions in busy urban environments, partly due to the lack of large-scale labeled ground-truth annotations for learning occlusion.