However, it does show the principle, and the model works. … and it works! It even recognizes models that were not in the training set - like the dusty old DC3-Dakota and the yellow Tiger Moth! It’s not a very good model it does not always recognize planes, but then again - 33 pictures are way too few for a good dataset. I still have the pictures from the last experiment, so why not re-use them? I trained a YoloV8 model on 33 pictures, adapted my code from April a little, deployed it on my HoloLens 2 …. Once again, I returned to my very first Computer Vision experiment: recognizing the model aircraft I made as a teenager. I really wanted to do that with custom data. In April, I described my experiments with computer vision, using a HoloLens 2, the Unity Barracuda inference engine, and an off-the-shelf YoloV7 model to recognize some objects in space. HoloLens AI - training a YoloV8 model locally on custom pictures to recognize objects in 3D space
