EfficientDet with Intel OpenVINO

This repository demonstrates how to convert AutoML EfficientDet to OpenVINO IR.

Follow the steps from .github/workflows/main.yml to convert your model. For public models, download IRs from GitHub Actions

How to convert model

1. Freeze graph

   cd automl/efficientdet
   python3 model_inspect.py --runmode=saved_model --model_name=efficientdet-d4 --ckpt_path=efficientdet-d4 --saved_model_dir=savedmodeldir

2. Create IR

   git clone https://github.com/openvinotoolkit/openvino --depth 1
   
   python3 openvino/model-optimizer/mo.py \
     --input_model efficientdet-d4.pb \
     --transformations_config openvino/model-optimizer/extensions/front/tf/automl_efficientdet.json \
     --input_shape "[1, 1024, 1024, 3]"

   find resolution of your model at https://github.com/google/automl/blob/master/efficientdet/hparams_config.py

   automl_efficientdet.json contains topology hyper-parameters

3. Validate model comparing accuracy with an original frozen TensorFlow graph

   python3 scripts/validate.py --version d4 --width 1024 --height 1024