====== Workshop Notes ====== ===== Preparation / Prerequisits ===== * Download ... * Install ... * Print ... ===== Introduction ===== * The work of the EOLab Team -> Current state of development * Image Classification * Object detection * Mini drones with OD ===== Hands On ===== * Connect SNAP to the server in Nvidia Jetson * Image classification game * Object Detection ?? ===== Reflection ===== ====== Main Achievements (internal discussion) ====== ===== SNAP! and Mini-Drone (Harley, 3 mins, live, with Alonzo pilot) ===== * Tello SNAP Backend (Javascript backend, communication software interface, Wifi, client, binding to IP address), URL, eolab.de github * One drone has a default IP, it is in "station" mode (the drone is AP, AP mode), 192.168.10.1 * Tello AP mode (client to Wifi), necessary for more than one drone in network and/or interaction with Jetson * Tello SNAP! category (collection of SNAP! Javascript blocks), websocket interaction with the interface talking to the drone * https://wiki.eolab.de/doku.php?id=drones:mini_drones:snap_tello ===== Jetbot and Object Detection with SNAP! (Ali, 3 mins, with Alonzo driver) ===== * Object follower * Jetbot Camera * SNAP! is running remotely, could be running on Jetbot * DetectNet (SSD-MobileNet V2, CoCo Dataset, 91 Classes) * Closed Loop Control * TODO: Short video! ===== Object Detection: Follow an Object with Drone (Ilgar, 3 mins, with Alonzo pilot) ===== * Based on Harley's presentation on Tello SNAP! interaction * New aspect: Object detection, Jetson * Challenges * Video stream from Tello drone to SNAP! (25 fps) * Video stream from SNAP! to Jetson (extacting stage in base64 format, send message, wait response, sequential, 7 fps) * Receive response from Jetson to SNAP! (bounding box, class label, coordinate transformation to the stage) * Frame rate incl. analysis is 7 fps * Problem (not serious): Realtime delay (latency) within the Tello drone video stream! * TODO: Short video!