Computer Vision

Training a YOLOv4 network for drone detection and implementation on an NVIDIA Jetson TX2 installed on a UAV

Build video dataset of flying drones for drone detection network

Eye super-resolution from synthetic data for improved gaze estimation in unconstrained settings

Estimate vehicle's position with respect to cones by using a thermal camera

Building models to improve transfer learning techniques for breast cancer detection using small data sets of mammograms and US scans

Design and train an algorithm for removing atmospheric turbulence through deep learning

Segmentation of corals using data augmentation and training of existing ConvNet

Anti-spoofing by using a dual camera system to capture image depth from two different angles

This project presents structured light method for calculating object surface using projected light stripes

We present an improved algorithm for horizon calculation for USV from video taking into account the boat's oscillations due to waves.

Collect data base and train a CNN to detect drones in images

Control TELLO drone to automatically follow human face

Real time cone detection and mapping solution using LiDAR & IMU integration for Technion FORMULA team.

Control a single drone using a ground stereo camera system

Create data set and train Resnet for smoke classification in images

Monitor traffic conditions under hard conditions at the exit of Tel Aviv HaBima parking lot

Recognition and identification of obstacles around USV using YOLO network

Autonomous drone localization and navigation from ground station camera

Analysis of a hidden scene by Time of Flight (ToF) using MITSUBA simulation tool

Improve the STGV algorithm for depth recovery using state of the art Deep Learning based segmentation techniques

Demonstration of global and direct illumination experiment using Mitsuba simulation.

Use morphological features for classification of benign or malignant tumors in mammogram images

Find and blur license plate numbers in surveillance video for privacy protection

Perform autonomous object tracking with hacked Nano Drone

LSD-SLAM and ORB-SLAM was implemented on a Turtlebot robot working under ROS for mapping and exploration purpuses.

IoT System to be in stalled on Smart Street Lights based on Raspberry Pi and Amazon web service.

Prototype of a moving robot arm system with capability of going back exactly to the same position as it started.

Using CNN neural network for classification of human \ non human objects in thermal images

Improve the vertical density of point clouds captured by LiDAR 16 depth camera.

The majority of car accidents which involve pedestrians occur due to Moving violations cause by the driver, therefor our main goal in this project was to detect pedestrians near pedestrian crossings in low visibility conditions in order to alert drivers.

Finding and navigating a path in a non-deterministic, 2D environment under probabilistic constraints.

Recognize and track pedestrians at crosswalks using Histogram of Oriented Gradients.

Increasing demand for smart pedestrian crossing control all over the world. Haifa municipally is looking for a solution in collaboration with the Technion.

In the project, we will build an exhibition for the illusion of a three dimension in an interactive way.

Analyze a Backgammon Game Board State from an RGB Image Create an Android App Create an easy-to-use wrapper for the GNU-BG algorithm Get Hands-On Experience Creating Image Process Algorithms

ThermTrack is an Android application that enables to track objects with a low-cost thermal camera of type Thermapp according to temperature range and basic image features. Image processing features like blob detection and object area, circularity, convexity etc. were taken from OpenCV libraries. The original goal was tracking pedestrians, pedestrian tracking can be utilized by the operator by choosing suitable parameters of the available features. The development was done using...

Traffic collision cost the life of hundreds per year, if you do not include the wounded. There are many roads which not contain traffic lights, despite considered dangerous for pedestrian. The main goal of this project is to develop a real-time system which warns when pedestrian approach or crossing the road, moreover we tried to implement the system on portable device in order to be stationed in junctions with no...

Implementation of sparsity based image processing methods on naturel texture image with stochastic features for the purpose of noise filtering and reconstruction of blur.

In recent years drones became more accessible and we would like to explore the magnificent world of drones and autonomous systems. We would like to use the drone in order to accomplish basic mission, and create an infrastructure for next to come

Heart Rate (HR) is not constant , it is modulated by different systems , which cause fluctuations in HR that are called Heart Rate Variability.