At Fives CortX we enhance industrial data to optimize your operations and maintenance.Through the implementation of our end-to-end technology stack, we connect your assets and build predictive models, to tackle your problems with:- predictive maintenance- real-time quality models for an optimized production steering- optimization of assets performance- remaining lifetime estimationOur technology integrates machine vision capabilities that tackle your quality inspection, live process monitoring, and traceability problems.Reporting to the CEO, I'm in charge of a dedicated computer vision team that develops a machine vision platform that handles camera acquisition, data fusion, image processing, and real-time data visualization.𝗩𝗶𝘀𝗶𝗼𝗻 𝘀𝘆𝘀𝘁𝗲𝗺 𝘀𝗲𝗹𝗲𝗰𝘁𝗶𝗼𝗻 & 𝗶𝗻𝘁𝗲𝗴𝗿𝗮𝘁𝗶𝗼𝗻- GigE/USB3 vision cameras- Profilometers/3D cameras- Lens and light selection- 𝗦𝗼𝗳𝘁𝘄𝗮𝗿𝗲 𝗦𝘁𝗮𝗰𝗸: C++, MVTec Halcon, Stemmer-Imaging CVB, Imaging Source SDK𝗗𝗲𝗲𝗽 𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴 𝗳𝗼𝗿 𝗱𝗲𝗳𝗲𝗰𝘁 𝗱𝗲𝘁𝗲𝗰𝘁𝗶𝗼𝗻 - 𝗦𝗼𝗳𝘁𝘄𝗮𝗿𝗲 𝗦𝘁𝗮𝗰𝗸: Python, Jupyter Notebooks, TensorFlow𝗔𝗱𝘃𝗮𝗻𝗰𝗲𝗱 𝗩𝗶𝘀𝘂𝗮𝗹𝗶𝘀𝗮𝘁𝗶𝗼𝗻 𝗮𝗻𝗱 𝘃𝗶𝘀𝗶𝗼𝗻-𝗼𝗿𝗶𝗲𝗻𝘁𝗲𝗱 𝗪𝗲𝗯 𝗔𝗽𝗽𝘀- 𝗦𝗼𝗳𝘁𝘄𝗮𝗿𝗲 𝗦𝘁𝗮𝗰𝗸: - Advanced Visualisation: Plotly/Dash, Grafana, PaperJS - Front End: TypeScript/JS-ES6, Angular, React - Backend: Golang, NodeJS Express, Node addons with N-API C++ integration, - Databases: Influxdb; SQLite, MySQL Knexjs - Desktop app: Electron 𝗣𝗿𝗼𝗷𝗲𝗰𝘁 𝗠𝗮𝗻𝗮𝗴𝗲𝗺𝗲𝗻𝘁- Prestudy and pre-sells- Product roadmap management- Agile/Scrum methodology with Atlassian Jira

𝗗𝗲𝘃𝗲𝗹𝗼𝗽𝗺𝗲𝗻𝘁 𝗼𝗳 𝗮 𝘀𝗺𝗮𝗿𝘁 𝗲𝗺𝗯𝗲𝗱𝗱𝗲𝗱 𝗰𝗮𝗺𝗲𝗿𝗮𝘀 𝘀𝗼𝗹𝘂𝘁𝗶𝗼𝗻- Real-time multiple object tracking algorithms- Data-fusion algorithms for the coordination of a network of smart cameras - Cameras aware of the industrial environment and able to use contextual information for online inference- Graphical tools for the configuration of large network cameras- Visualization of the meta-data generated by the smart cameras- 𝗩𝗶𝘀𝗶𝗼𝗻 𝗮𝗽𝗽𝗿𝗼𝗮𝗰𝗵𝗲𝘀: Multiple object tracking, object detection, background subtraction, occlusion handling, Kalman Filtering, Object Re-Identification.- 𝗦𝗼𝗳𝘁𝘄𝗮𝗿𝗲 𝘀𝘁𝗮𝗰𝗸: C++11/Python, OpenCV, GStreamer, Qt, ZeroMQ Boost, Yocto- 𝗛𝗮𝗿𝗱𝘄𝗮𝗿𝗲: NXP® i.mx 6 ARM, Nvidia Jetson board, MIPI/CSI Cameras𝗜𝗻𝗱𝘂𝘀𝘁𝗿𝗶𝗮𝗹 𝗮𝗽𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻𝘀 :- Vision module of an object picking robot -- Fusion of 2D/3D sensors -- Object segmentation based on Deep Learning approaches (Mask-RCNN/Yolo) -- Semi-automatic annotation approaches for large dataset generation -- 𝗦𝗼𝗳𝘁𝘄𝗮𝗿𝗲 𝘀𝘁𝗮𝗰𝗸: Python/TensorFlow – C++11/PCL- Deployment of a large network of smart cameras for the control of an industrial facility 𝗥𝗲𝘀𝘂𝗹𝘁𝘀: Less than 1 in 1000 tracking errors- Camera-based optimization of an industrial process 𝗥𝗲𝘀𝘂𝗹𝘁𝘀: 40% improvement with the camera-based solution compared to standard process

The main subject of my thesis is multiple object tracking using a network of smart cameras deployed in an industrial application. I worked on the development of smart cameras network able to automatically extract and exploit contextual information collected from an industrial environment in order to perform robust and real-time tracking. The goal of the camera-based tracking is the control of an industrial facilityMy thesis was under the supervision of Prof. Serge Miguet and Mihaela Scuturici .Publications: Adel Benamara, Serge Miguet & Mihaela Scuturici. « Multiple Object Tracking on SmartEmbedded Camera For Automated Conveying Systems ». 11 th ACM/IEEE InternationalConference on Distributed Smart Cameras (ICDSC’17), Stanford University, CA, USA, 5-7 Sep,2017.Adel Benamara, Serge Miguet & Mihaela Scuturici (2016). « Real-time Multi-Object Tracking with Occlusion and Stationary Objects Handling for Conveying Systems ». 12th International Symposium on Visual Computing (ISVC'16), 14 décembre 2016, Las Vegas, NV (États-Unis).