Ajinkya Khoche

Lindstedtsvägen 24

Stockholm, Sweden

I’m a PhD candidate at the division of Robotics Perception and Learning at KTH, supervised by Prof. Patric Jensfelt and Dr. Sina Sharif Mansouri. My thesis focuses on long-range perception for autonomous trucking. My work is part of the ProSense project, supported by Vinnova.

During my PhD thesis, I have explored topics in scene perception, including object detection, segmentation and scene flow. I am also interested in 3D forecasting, scene rendering and world models. I am passionate about developing deployable and scalable methods that are robust to adverse weather and long-tail. I have experience in working with various sensors like LiDAR, cameras, and radar.

Previously, I graduated with M.Sc. in Systems, Controls and Robotics from KTH. My M.Sc. thesis explored depth estimation supervised using dense correspondences between image and SLAM-processed LiDAR under supervision of Dr. Mentar Mahmoudi at Audi Electronic Ventures, Ingolstad. I also hold B.Tech. + M.Tech, in Mechanical Engineering from Indian Institute of Technology, Kharagpur.

news

Jan 31, 2025	Our paper SSF: Sparse Long-Range Scene Flow for Autonomous Driving is accepted to ICRA 2025. Source code is released here. See you in Atlanta!
Jun 02, 2024	Reporting from beautiful Jeju in South Korea. I’m here to present our paper Towards long-range 3d object detection for autonomous vehicles, accepted in IV 2024. Later during the month I’ll be at ECC 2024 in Stockholm to present another accepted paper Addressing Data Annotation Challenges in Multiple Sensors: A Solution for Scania Collected Datasets
Jun 06, 2022	Semantic 3d grid maps for autonomous driving is accepted to ITSC 2022!

selected publications

ICRA 2025

SSF: Sparse Long-Range Scene Flow for Autonomous Driving

Ajinkya Khoche, Qingwen Zhang, Laura Pereira Sanchez, and 3 more authors

arXiv preprint arXiv:2501.17821, 2025

arXiv Bib Code

@article{khoche2025ssf,
  title = {SSF: Sparse Long-Range Scene Flow for Autonomous Driving},
  author = {Khoche, Ajinkya and Zhang, Qingwen and Sanchez, Laura Pereira and Asefaw, Aron and Mansouri, Sina Sharif and Jensfelt, Patric},
  journal = {arXiv preprint arXiv:2501.17821},
  year = {2025},
}

ECC 2024

Addressing Data Annotation Challenges in Multiple Sensors: A Solution for Scania Collected Datasets

Ajinkya Khoche, Aron Asefaw, Alejandro González, and 3 more authors

In 2024 European Control Conference (ECC), 2024

arXiv Bib

@inproceedings{khoche2024addressing,
  title = {Addressing Data Annotation Challenges in Multiple Sensors: A Solution for Scania Collected Datasets},
  author = {Khoche, Ajinkya and Asefaw, Aron and Gonz{\'a}lez, Alejandro and Timus, Bogdan and Mansouri, Sina Sharif and Jensfelt, Patric},
  booktitle = {2024 European Control Conference (ECC)},
  pages = {1032--1038},
  year = {2024},
  organization = {IEEE},
}

IV 2024

Towards long-range 3d object detection for autonomous vehicles

Ajinkya Khoche, Laura Pereira Sánchez, Nazre Batool, and 2 more authors

In 2024 IEEE Intelligent Vehicles Symposium (IV), 2024

arXiv Bib

@inproceedings{khoche2024towards,
  title = {Towards long-range 3d object detection for autonomous vehicles},
  author = {Khoche, Ajinkya and S{\'a}nchez, Laura Pereira and Batool, Nazre and Mansouri, Sina Sharif and Jensfelt, Patric},
  booktitle = {2024 IEEE Intelligent Vehicles Symposium (IV)},
  pages = {2206--2212},
  year = {2024},
  organization = {IEEE},
}

IV 2022

Semantic 3d grid maps for autonomous driving

Ajinkya Khoche, Maciej K Wozniak, Daniel Duberg, and 1 more author

In 2022 IEEE 25th International Conference on Intelligent Transportation Systems (ITSC), 2022

arXiv Bib

@inproceedings{khoche2022semantic,
  title = {Semantic 3d grid maps for autonomous driving},
  author = {Khoche, Ajinkya and Wozniak, Maciej K and Duberg, Daniel and Jensfelt, Patric},
  booktitle = {2022 IEEE 25th International Conference on Intelligent Transportation Systems (ITSC)},
  pages = {2681--2688},
  year = {2022},
  organization = {IEEE},
}

M.Sc. Thesis
Depth Estimation from Images using Dense Camera-Lidar Correspondences and Deep Learning

Ajinkya Khoche

KTH, School of Electrical Engineering and Computer Science (EECS), 2020

Abs Bib PDF

Depth estimation from 2D images is a fundamental problem in Computer Vision, and is increasingly becoming an important topic for Autonomous Driving. A lot of research is driven by innovations in Convolutional Neural Networks, which efficiently encode low as well as high level image features and are able to fuse them to find accurate pixel correspondences and learn the scale of the objects. Current state-of-the-art deep learning models employ a semi-supervised learning approach, which is a combination of unsupervised and supervised learning. Most of the research community relies on the KITTI datasets for benchmarking of results. But the training performance is known to be limited by the sparseness of the lidar ground truth as well as lack of training data. In this thesis, multiple stereo datasets with increasingly denser depth maps are generated on the corpus of driving data collected at the Audi Electronics Venture GmbH. In this regard, a methodology is presented to obtain an accurate and dense registration between the camera and lidar sensors. Approaches are also outlined to rectify the stereo image datasets and filter the depth maps. Keeping the architecture fixed, a monocular and a stereo depth estimation network each are trained on these datasets and their performances are compared to other networks reported in literature. The results are competitive, with the stereo network exceeding the state-of-the-art accuracy. More work is needed though to establish the influence of increasing depth density on depth estimation performance. The proposed method forms a solid platform for pushing the envelope of depth estimation research as well as other application areas critical to autonomous driving.
@mastersthesis{Khoche1417257, author = {Khoche, Ajinkya}, institution = {KTH, School of Electrical Engineering and Computer Science (EECS)}, school = {KTH, School of Electrical Engineering and Computer Science (EECS)}, title = {Depth Estimation from Images using Dense Camera-Lidar Correspondences and Deep Learning}, series = {TRITA-EECS-EX}, number = {2020:12}, year = {2020}, }