Real-time UAV & swarm detection on edge GPUs
YOLO- and RT-DETR-based detectors running in real time on NVIDIA Jetson Orin, with INT8 quantization-aware training and an error taxonomy that targets data collection.
I am a machine-learning and computer-vision researcher building real-time perception that runs on the edge. My work spans object detection, SLAM, and model compression, and I take systems from research prototype to on-device deployment. I hold a PhD in computational neuroscience and a machine-learning post-doc, with research published across neuroscience and vision.
Research interests: real-time object detection, visual-inertial SLAM, quantization & pruning for edge inference, multi-camera calibration, dataset curation, and Auto-ML.
Projects
Perception for autonomous mobile robots
Obstacle detection with wide-area mapping and localization, with models compressed via quantization and pruning to hold real-time latency on-device.
Multi-camera perception & auto-calibration
Autonomous multi-camera perception with real-time auto-calibration that keeps stereo rigs aligned in the field, without manual resets.
Classification, image-quality & face models
Core vision models for image classification, multi-label tagging, image-quality enhancement, and face toonification (cartoon-style stylization).
Stereo-IMU fusion SLAM
Stereo + IMU visual-inertial SLAM tuned to run in real time within the compute budget of mobile chipsets.
Stereo depth & real-time tracking
Stereo disparity and depth segmentation with object tracking and real-time head detection on embedded vision SoCs.
Blog
-
What CVPR 2026 Says About Production Computer Vision
June 2026
A practical review of what can ship now, what is near, and where research still outruns deployment.
-
The Pipeline Is Moving Inside the Model
June 2026
Software 2.0 replaced vision algorithms. The next transition is replacing the interfaces between them.
-
The Map of CVPR 2026
June 2026
A data-driven tour of all 2,672 CVPR 2026 papers — an interactive map of the conference, the topics that dominate it, the ideas everyone is building on, and how the field names its work.
-
Starting a notebook on edge perception
June 2026
Why I’m writing here, and the kind of notes to expect — short, practical pieces on real-time detection, SLAM, and squeezing models onto edge hardware.
Education
-
Post-Doctoral Fellow, Machine Learning 2012–2013Weizmann Institute of Science
-
PhD, Computational Neuroscience 2007–2011Ben-Gurion University of the Negev
-
MSc, Computational Biology 2003–2006Tel-Aviv University
-
BSc, Mathematical & Life Sciences 2000–2003Tel-Aviv University
Selected publications
- A. Weissbrod, A. Shapiro, G. Vasserman, L. Edry, M. Dayan, A. Yitzhaky, L. Hertzberg, O. Feinerman, T. Kimchi. Automated long-term tracking and social behavioural phenotyping of animal colonies within a semi-natural environment. Nature Communications, 2013.
- G. Vasserman, E. Schneidman, R. Segev. Adaptive color contrast coding in the retina efficiently matches natural scene statistics. PLOS ONE, 2013.
- M. Wacht Katz, Z. Abramsky, B. P. Kotler, M. L. Rosenzweig, O. Altstein, G. Vasserman. Optimal foraging of Little Egrets in a patchy environment. The American Naturalist, 2013.
- A. Ben-Simon, O. Ben-Shahar, G. Vasserman, M. Ben-Tov, R. Segev. Visual acuity in the archer fish: behavior, anatomy, and neurophysiology. Journal of Vision, 2012.
- A. Ben-Simon, O. Ben-Shahar, G. Vasserman, R. Segev. Predictive saccade in the absence of smooth pursuit: interception of moving targets in the archer fish. Journal of Experimental Biology, 2012.
- Z. Agur, O. U. Kirnasovsky, G. Vasserman, L. Tencer-Hershkowicz, Y. Kogan, H. Harrison, et al. Dickkopf1 regulates fate decision and drives breast cancer stem cells to differentiation: an experimentally supported mathematical model. PLOS ONE, 2011.
- G. Vasserman, M. Shamir, R. Segev. Coding “what” and “when” in the archer fish retina. PLoS Computational Biology, 2010.
- A. Plotkin, E. Paperno, G. Vasserman, R. Segev. Magnetic tracking of eye motion in small, fast-moving animals. IEEE Transactions on Magnetics, 2008.
- G. Vasserman, L. G. Magal, J. Shepshelovich, E. Elifaz, K. Hirschberg. Processing of VSVG protein is not a rate-limiting step for its efflux from the Golgi complex. Biochem. Biophys. Res. Commun., 2006.
Patents
- Method for automatic behavioral phenotyping. Yeda — US 2014/0167958.
- System and method for multiple feature detection and analysis of a rotating tire. Neomatix.
- Storing data retrieved from different sensors for generating a 3-D image. Inuitive.