Predict the wave. Perfect the ride 

Sam Sotil, Eliya Zakay, Shalev Atsis

GitHub

A computer-vision system that automatically analyzes surfing footage to estimate key wave properties such as height, period, shape, speed, and breaking dynamics.

See differences. Ensure alignment 

Liran Aichnboim, Afik Aharon, Adi Haim

GitHub

A vision- and geometry-based system that compares two versions of the same floor plans to identify structural differences, inconsistencies, or missing elements. The model aligns layouts, detects deviations in walls, rooms, openings, and dimensions, and highlights mismatches with high precision.

Understand engagement, not just attendance. 

Ofir Duek, Rotem Aloni, Aviv Meir

GitHub

A video-only analytics system that monitors student behavior during remote Zoom-style sessions by interpreting visual cues, including gaze direction, facial engagement, posture, movement patterns, and interaction gestures. It estimates attention levels, focus fluctuations, signs of distraction, and overall engagement dynamics without relying on audio or chat signals.

Know the cause. Act with precision 

Adir Boccara, Aharoni Cohen, Eden Charkachi

GitHub

A vision-based diagnostic system that analyzes leaf and plant imagery to distinguish between chemical damage and biological threats such as fungi, pests, or pathogens. By learning subtle visual patterns: discoloration types, lesion shapes, spread dynamics, and texture changes, the model identifies the underlying cause of crop stress with high reliability.  

Spot risks before they become accidents 

Dan Kuenkas, Nadav Golche

GitHub

A video-based safety monitoring system that automatically detects construction workers who are not wearing protective helmets. Using real-time object and attribute recognition, the model identifies workers, classifies helmet presence, and tracks individuals across the site.  

See hazards before they reach small hands. 

Svetlana Gavris, Semen Ostrovsky, Gunko Daniel

GitHub

A vision-based system that detects objects in a child’s environment that may pose safety risks, such as sharp tools, choking hazards, toxic items, or hot surfaces. The model analyzes indoor scenes to identify dangerous objects, their proximity to a child, and the context in which they appear. 

See through contamination. Detect with confidence 

Yuval Rubin, Shahaf Levi, Guy Yogev

GitHub

A vision-based ADAS module that maintains reliable object detection even when mud, dust, rain droplets, snow, or glare partially obstruct the camera lens. The system identifies the type and severity of contamination, compensates for degraded image regions, and preserves detection accuracy for vehicles, pedestrians, road signs, and hazards. 

Spot every threat. Protect every landing. 

Yuri Matyash,Ofek Igud, Hadar Lavsky

GitHub

A vision-based system that detects foreign objects and surface contamination on airport runways, including debris, tools, wildlife remnants, fluids, and particulate hazards. The model analyzes high-resolution imagery from ground vehicles, fixed cameras, or drones to identify abnormalities, classify risk levels, and localize them precisely along the runway.  

Catch defects before they reach the customer. 

Amit Wagensberg, Yaniv Hananis, Ori Zarfati

GitHub

A vision-based inspection system for TV production lines that identifies physical defects such as dead pixels, scratches, pressure marks, misaligned panels, edge cracks, and assembly irregularities. Using high-resolution imaging and precise geometric analysis, the system detects subtle anomalies in real time as units move along the manufacturing line. 

Understand shopper actions as they happen

Bar Sberro, Arbel Koren, Noy Leibovitch

GitHub

A video-based retail analytics system that detects when shoppers pick up, hold, or inspect products on shelves. By analyzing body posture, hand–object interactions, and movement patterns, the model identifies moments of product engagement and distinguishes them from casual browsing.

Find open spaces with precise vision. 

Lior Yanwo, Kiril Litvin, Stas Shuvaev

GitHub

A vision-based ADAS module that automatically detects empty street parking slots using real-time camera input. The system analyzes road edges, markings, curb geometry, and vehicle spacing to identify available parking spaces while the car is in motion. It distinguishes valid spots from blocked or partially occupied areas and provides accurate localization for driver assistance or autonomous parking workflows 

See through the weather. Detect what matters.

Rachel Sade, Yuval Pery, Shaked Horesh

GitHub

A vision-based detection system designed for drones operating under adverse weather conditions such as rain, fog, snow, haze, or strong winds. The model compensates for degraded visibility and motion artifacts to identify people, vehicles, and critical objects on land or at sea.

Find lives in the waves

Shalev Cohen,Tomer Atia, Noam Hadad

GitHub

A vision-based detection system that identifies humans in turbulent or stormy sea conditions using drone, vessel-mounted, or coastal cameras. The model analyzes dynamic water patterns, foam, occlusions, and low-contrast visuals to detect swimmers, survivors, or overboard individuals even in rough seas.

Measure quality. Elevate professionalism 

Daniel Ziv,Daniel Buts

GitHub

A vision-based system that evaluates the quality of professional headshot images by analyzing factors such as lighting, sharpness, facial framing, background cleanliness, pose aesthetics, color balance, and overall visual appeal.

See early. Treat smarter.

Ron Noiman, Lin Schneider, Reut Sasson

GitHub

A vision-based system that detects and classifies facial skin problems such as acne, redness, pigmentation, dryness, lesions, and irritation. Using high-resolution imagery and dermatology-informed features, the model identifies affected regions, categorizes condition types, and provides severity assessments.

Perfect your form. Prevent injury.

May Eden, Rotem Pasharel, Norfar Hatam

GitHub

A vision-based system that analyzes exercise movements to detect mistakes in body position, alignment, and motion patterns. Using pose estimation and biomechanical cues, the model identifies deviations from correct technique, including improper angles, unstable posture, and asymmetrical movement. 

Spot the fake. Trust the pair. 

Amit Mitzmacher, Tal Mitzmacher, Dani Gorodnitsky

GitHub

A vision-based system that authenticates branded footwear by analyzing detailed visual cues such as stitching quality, logo placement, material texture, outsole patterns, and shape geometry. By comparing captured images to learned models of genuine products, it identifies inconsistencies that signal counterfeit items. 

Early risk assessment of skin lesions from images. 

Shir Molakandove, Afik Haviv, Romi yosef

GitHub

MoleGuard is a computer-vision project that classifies skin moles by malignancy risk using dermoscopic or standard clinical images. The system analyzes visual cues, including asymmetry, border irregularity, color variation, and texture, to distinguish benign lesions from potentially malignant lesions. 

Real-time weapon detection from body-worn cameras. 

Aviv Heller, Afik Suisa, Hen Golyan

GitHub

SentinelCam is a computer-vision system designed to detect firearms and bladed weapons in body-camera imagery automatically. It operates under challenging real-world conditions, including motion blur, partial occlusion, low light, and extreme viewpoints, providing timely alerts for operational awareness and enabling efficient indexing of recordings for post-event analysis in security and public-safety scenarios.

Abandoned object detection in crowded public spaces. 

Israel Peled, Natan Shick, Orel Cohen

GitHub

ClearZone is a computer-vision system for detecting unattended and abandoned objects in crowded, cluttered environments, including transportation hubs, streets, and public venues. The system analyzes long-term scene dynamics, human–object interactions, and occlusions to distinguish truly abandoned items from transient clutter, enabling early alerts for public safety monitoring and effective post-incident investigation.

Recognizing spaces, not appearances. 

Arkadi Doktorovich, Elia Meerson

GitHub

SameRoom identifies whether two images depict the same physical room despite changes in viewpoint, lighting, decor, objects, or people by learning room-level invariants rather than relying on surface-level visual similarity.