TFlite implementations from https://github.com/sravansenthiln1/armnn_tflite/ adapted to run on Verisilicon's Vivante NPU hardware platform.
Compatible with VIM3/3L, on the 4.9 and 5.15 kernel.
sudo apt-get install python3-pip
pip3 install numpy pillow opencv-python librosa
pip3 install --extra-index-url https://google-coral.github.io/py-repo/ tflite_runtime
sudo ln /usr/lib/libOpenVX.so /usr/lib/libOpenVX.so.1
If you are running on VIM3:
sudo cp libs/VIM3/libtim-vx.so /usr/lib/aarch64-linux-gnu/
If you are running on VIM3L:
sudo cp libs/VIM3L/libtim-vx.so /usr/lib/
model | description |
---|---|
Sine Model | Basic Neural network TFLite model |
Digit recognize Model | Digit recognization model |
model | description |
---|---|
Audio classifier Model | Audio classifier model |
model | description |
---|---|
Mobilenet v1 Model | Mobilenet v1 image classification model |
YOLOv8n Model | YOLOv8n image detection model |
Auto crop Model | Automatic document crop model |