Jesper Haahr Christensen

Jesper Haahr Christensen

PhD Student

Technical University of Denmark

Short Bio

I am a Ph.D. candidate under the supervision of Ole Ravn in the Department of Electrical Engineering at the Technical University of Denmark. During my Ph.D., I am also spending time and taking part in ongoing research in the Computer Vision Group of Stella Yu, and under the supervision of Sascha Hornauer, at the International Computer Science Institute at UC Berkeley. I received my B.Eng. and M.Sc. degree from the Department of Electrical Engineering at the Technical University of Denmark, and spend a semester abroad at Queensland University of Technology.

My research is in Computer Vision and Machine/Deep Learning applications for perception, recognition, and data analysis using multiple sensor modalities. Currently, my focus is within the maritime domain.

Talks

Applications of Deep Learning in Autonomous Underwater Vehicles (AUVs)
Nov 8, 2019 ATLAS ELEKTRONIK, Bremen, Germany

Publications

Jesper Haahr Christensen
May 2021 In Hydrographische Nachrichten, May 2021
Automatic Boulder Identification in Side-Scan Sonar

Automatic Boulder Identification in Side-Scan Sonar

Side-scan data is automatically processed to estimate boulder positions and generate metadata.

PDF DOI
Jesper Haahr Christensen, Lars Valdemar Mogensen, Ole Ravn
May 2021 For IFAC CAMS, 2021
Side-Scan Sonar Imaging: Real-Time Acoustic Streaming

Side-Scan Sonar Imaging: Real-Time Acoustic Streaming

Live side-scan sonar stream is compressed to below 1.87 kbit/s and reconstructed to high-quality estimations on the receiving end.

PDF DOI
Jesper Haahr Christensen, Marco Jacobi, Martin Clemmensen Rotne, Morten Soede Nielsen, Max Abildgaard, Claus Eriksen, Lars Valdemar Mogensen
September 2020 In IEEE OES Autonomous Underwater Vehicle (AUV) Symposium, 2020
SeaShark: Towards a Modular Multi-Purpose Man-Portable AUV

SeaShark: Towards a Modular Multi-Purpose Man-Portable AUV

We present the SeaShark AUV; an in-house developed small, easily configurable, one-man-operational autonomous underwater vehicle (AUV)

PDF Video DOI
Jesper Haahr Christensen, Lars Valdemar Mogensen, Ole Ravn
September 2020 In IEEE OES OCEANS, 2020
Single Image Super-Resolution for Domain-Specific Ultra-Low Bandwidth Image Transmission

Single Image Super-Resolution for Domain-Specific Ultra-Low Bandwidth Image Transmission

Original underwater images are downsampled to a low-resolution low-size thumbnail to be transmitted over acoustics. We learn to reconstruct the transmitted low-res image to high-res super-resolved version.

PDF Video DOI
Jesper Haahr Christensen, Lars Valdemar Mogensen, Ole Ravn
July 2020 In International Federation of Automatic Control (IFAC), 2020
Deep Learning based Segmentation of Fish in Noisy Forward Looking MBES Images

Deep Learning based Segmentation of Fish in Noisy Forward Looking MBES Images

We predict binary segmentation masks on sonar images to detect fishes.

PDF Video DOI
Jesper Haahr Christensen, Sascha Hornauer, Stella Yu
June 2020 In Sight and Sight Workshop, Conference on Computer Vision and Pattern Recognition (CVPR), 2020
BatVision with GCC-PHAT Features for Better Sound to Vision Predictions

BatVision with GCC-PHAT Features for Better Sound to Vision Predictions

We improve our previous BatVision sound-to-vision model.

PDF Video DOI
Jesper Haahr Christensen, Sascha Hornauer, Stella Yu
June 2020 In International Conference on Robotics and Automation (ICRA), 2020
BatVision: Learning to See 3D Spatial Layout with Two Ears

BatVision: Learning to See 3D Spatial Layout with Two Ears

We predict 2D and 3D spatial layout from binaural echolocation.

PDF DOI
Jesper Haahr Christensen, Lars Valdemar Mogensen, Roberto Galeazzi, Jens Christian Andersen
November 2018 In 2018 IEEE/OES Autonomous Underwater Vehicle Workshop (AUV)
Detection, Localization and Classification of Fish and Fish Species in Poor Conditions using Convolutional Neural Networks

Detection, Localization and Classification of Fish and Fish Species in Poor Conditions using Convolutional Neural Networks

We predict the location and specie of fish in poor underwater images.

DOI