Recent Publications

Recent publications by TensorTour team members in top Computer Vision and Machine Learning venues

TensorTour Researchers have extensive proven track record of high quality research in the top venues in Computer Vision, Pattern Recognition, Machine Learning and Robotics. Over the years they have served (and are currently serving) as official Reviewers and Area Chairs of many such prestigious venues. Listed below are some of the recent publications of TensorTour team members.

Beyond Mono to Binaural: Generating Binaural Audio from Mono Audio with Depth and Cross Modal Attention

Winter Conference on Applications of Computer Vision (WACV), 2022

We show that using depth maps in addition to the mono audio as well the scenes RGB images, improves the task of generating binaural audio. The depth maps induce the information about the relative distances of the different sound sources in the scene and allows the method to predict the binaural sound signal better.
CVF open access version

Beyond Image to Depth: Improving Depth Prediction using Echoes

Conference on Computer Vision and Pattern Recognition (CVPR), 2021

Inspired by echolocation, used by animals such as Dolphins and Bats, as well as by visually impaired people, we propose a novel method which uses echoes produced by the scene to help predict the depth of the scene.
ArXiv open access version Blog Post

Self Attention Guided Depth Completion using RGB and SparseLiDAR Point Clouds

Conference on Intelligent Robots and Systems (IROS), 2021

Given the RGB image of the scene and sparse point cloud of the scene, we present a novel attention based method to generate the dense depth map. The method uses geometric primitives and self attention mechanisms to obtain improved results on challenging public benchmarks.

Exploiting Local Geometry for Feature and Graph Construction for Better 3D Point Cloud Processing with Graph Neural Networks

International Conference on Robotics and Automation (ICRA), 2020

We propose simple yet effective improvements in point representations and local neighborhood graph construction within the general framework of graph neural networks (GNNs) for 3D point cloud processing.
ArXiv open access version