Using the Edge TPU pre-trained model.
Download SSD MoblieNet v1 coco or SSD MoblieNet v2 coco model.
See Coral Models.
$ python3 ./camouflage_picamera.py --model=<PATH_TO_edgetpu.tflite> --label=./label.txt
- Space key: toggle detection and camouflage.
- q key: End app.
- top_k: The maximum number of detected objects to return (DetectWithImage).
- threshold: Minimum confidence threshold for detected objects (DetectWithImage).
- width: Width of the frames in the camera stream.
- height: Height of the frames int the camera stream.
You can change the target object by changing the label.txt file.
Default is banana(40 banana).
Select the target object from the coco label(index and name). Or use the model you created.