This directory contains a C++ Gstreamer example for classification model to work with the Coral Dev Board. The app itself mendelcam is built from a provided docker image that fetches all required dependencies and build it. The docker build was tested on the Dev Board, however, it can can support k8, aach64, and armv7a.

1. If you haven't done so already, set up your board according to Get started with the Dev Board.

2. Cross-compile for aarch64 from your Linux desktop (do not run this on the Dev Board):

cd nativeapp
make DOCKER_TARGETS=mendelcam DOCKER_CPUS=aarch64 docker-build

3. Copy models and binaries to the board: First download model:

../download_models.sh

Then copy models and binaries to the board:

mdt push imagenet_labels.txt
mdt push mobilenet_v2_1.0_224_quant_edgetpu.tflite
mdt push out/aarch64/demo/mendelcam

3. Connect to the device:

mdt shell

4. To run: In the nativeapp directory on the target (don't forget mdt shell) run:

./mendelcam --model ~/mobilenet_v2_1.0_224_quant_edgetpu.tflite  --labels ~/imagenet_labels.txt

Resolve permission error:

mendel@neat-yarn:~$ ./mendelcam --model ~/mobilenet_v2_1.0_224_quant_edgetpu.tflite  --labels ~/imagenet_labels.txt
-bash: ./mendelcam: Permission denied
mendel@neat-yarn:~$ sudo chmod 770 mendelcam 

How to use with USB camera:

To use with USB camera build the app by changing viderosrc /dev/video0 to /dev/video-<usb-camera-source> here