I need to build a Computer Vision and Deep Learning framework on NVIDIA Jetson TX2 Xavier devices

The project will be articulated in 3 phases, the first we will create the core, the part of the basic computer vision, which includes the configuration of the cameras, calibration, alignment and image optimization operations. Then the Tracking, sensor fusion, identification and data reading functions The last phase with we will implement the Machine and Deep Learning functionalities Phase 1 dettail Configuring LAN / USB video streams collection with json configuration file where details are listed (seonsore size, fov, type, etc.) The camera can be RGB visual, thermal, stereoscopic Create a function that calls answers with the cameras connected at same time Distortion correction functions (calibration) Video alignment features (different focal lengths and resolutions) in overlay with possible transparence Video optimization for Track Thermal spot evidence on R.G. B Single element traking Recognition of some object (like yolo) Color recognition object Vehicle and license plate recognition (plate color brand) Traking rgb + thermal sensor fusion Stereoscopic Deep Map from stereo cameras in overlay on the RGB Integration with imu and gps and IMU Geo tag position of tracked element (trigonometry) The targhet of the framework are working on tx2 in a Unmanned vehicle UAV and on AUV Rover. In a first time the delopment and testing phase are executed on the ground. In a second moment, where the integration with the onboard sensors like GPS and IMU, for example, is complete all the tests will be done on a little UAV. The final purpose of the framework will be very broad, being used for research and development in a university, a part have targhet at the nature and safety and sorting of salvatic animals such as: Identification of waste and plastic material for removal Research, cataloging and sorting of wild animals for identification count and making safe (e.g. roe deer before haymaking) Then a part for industrial waste in the energy sector: Problem recognition and position reporting in photovoltaic systems Cataloging of electric pylons and identification of problems such as overheating of joints or broken insulators For these reasons there must be a reliable trigonometric algorithm of GeoTag