Install python 3+ dependencies.

tensorflow: 1.12.0
onnx: 1.6.0
tf2onnx: 1.5.5
onnxruntime: 1.1.0

Download TF object detection model trained on COCO dataset from the Model Zoo and convert it to onnx model.
For example download ssd_mobilenet_v1_coco_2018_01_28 and convert it to onnx model from saved_model.pb

python -m tf2onnx.convert 
  --opset 11 
  --fold_const 
  --saved-model ssd_mobilenet_v1_coco_2018_01_28/saved_model/ 
  --output ssd_mobilenet_v1_coco_2018_01_28/model.onnx

or from frozen_inference_graph.pb

python -m tf2onnx.convert 
  --opset 11 
  --fold_const 
  --graphdef ssd_mobilenet_v1_coco_2018_01_28/frozen_inference_graph.pb 
  --output ssd_mobilenet_v1_coco_2018_01_28/frozen.onnx 
  --inputs image_tensor:0 
  --outputs detection_boxes:0,detection_classes:0,detection_scores:0,num_detections:0

For run python and C# examples below download path with already-made ssd_mobilenet_v1_coco_2018_01_28 onnx models and move it to repository root folder.

Run python script object_detection_image_onnx.py to test converted onnx model.

Figure 1. Python example

Build C# source code and run application.

Figure 2. C# example

[1] TensorFlow detection model zoo.
[2] Tutorial: how to convert them to ONNX and run them under onnxruntime.
[3] Microsoft: ONNX Runtime C# API.