firekind / resnet-mediapipe

A simple mediapipe project that uses a resnet-18 model pretrained on imagenet in a mediapipe pipeline.

Resnet-18 using mediapipe

This is a simple mediapipe project that uses a resnet-18 model pretrained on imagenet in a mediapipe pipeline. The pretrained model was converted from a pytorch model to tflite, and then included in the pipeline.
The pytorch model was converted to tflite using this gist.

Build

to build the project, run

$ bazel build -c opt --copt -DEGL_NO_X11 resnet-mediapipe:resnet-mediapipe

and to run the project, run

$ ./bazel-bin/resnet-mediapipe/resnet-mediapipe --calculator_graph_config_file=./resnet-mediapipe/graphs/resnet.pbtxt --image_path=./images/input.jpg