convert tensorflow object detection models:

openvino_dir=/opt/intel/openvino mo_dir=${openvino_dir}/deployment_tools/model_optimizer model_dir=pwd data_type=FP16

python3 ${mo_dir}/mo_tf.py
--input_model=${model_dir}/frozen_inference_graph.pb
--output=detection_boxes,detection_scores,num_detections
--data_type=${data_type}
--tensorflow_use_custom_operations_config=${mo_dir}/extensions/front/tf/ssd_v2_support.json
--tensorflow_object_detection_api_pipeline_config=${model_dir}/pipeline.config
--input=image_tensor
--reverse_input_channels
--input_shape=[1,300,300,3]

python3 ${mo_dir}/mo_tf.py
--input_model=${model_dir}/frozen_inference_graph.pb
--data_type=${data_type}
--tensorflow_use_custom_operations_config=${mo_dir}/extensions/front/tf/mask_rcnn_support.json
--tensorflow_object_detection_api_pipeline_config=${model_dir}/pipeline.config
--reverse_input_channels
--input=image_tensor
--input_shape=[1,800,1365,3]