Iteration 0 for the Autonomous Drone Forge Project.
Runs a server locally, and uses mjpeg to stream camera input to that server. The camera used is an IMX219 Arducam at 640 x 480p resolution. The Pi is flashed with Raspberry Pi OS Lite (32-bit).
Reads directly from the local server via IP address and port into an Opencv2 cam. Opencv then uses a cascade.xml file generated by the OpenCV's HAAR Cascade trainer to check camera frames for cone objects.
A Python script uses the icrawler API to go through Google, collecting at most 1000 images dating back from 2010. These files are organized into positive and negative categories, which are then fed into the HAAR Cascade trainer to generate the .xml file.
OpenCV2
icrawler
Cascade Trainer GUI: https://amin-ahmadi.com/cascade-trainer-gui/