MACHINE LEARNING MONDAY – MobileNet V4 SSD Lite on Raspberry Pi 4 @adafruit @raspberry_pi @tensorflow #machinelearning #tinyML #raspberrypi

This week we’re building on last week’s Machine Learning project where we run the MobileNet v2 1000-object detector on the Raspberry Pi 4 + BrainCraft HAT (video). This time we’re running MobileNet V4 SSD Lite, which can do segmented detections. in this case it has only 90 objects it can detect but it can draw a box around the objects found. That’s displayed on the 1.5″ TFT on the HAT for human verification. This model’s not as fast as the non-segment object detector, about 1.5 fps. We might be able to speed that up by running TensorFlow Lite instead of full TF – we’ll work on that next!