Can I run a normal tensorflow object detection model on Jetson board without using TensorRT?

Question

Is it possible to run the inference graph on the jetson board without converting it into TensorRT format as mentioned in the github repo ?

Will we be able to run the Tensorflow Object Detection API without using TensorRT on the Jetson board?

Answer 1

• Install TF from Nvidia website for the right jetpack version.

• Coming to the object detection, execute first code block under Step 1, mentioned in this link

    git clone --quiet https://github.com/tensorflow/models.git
    
    apt-get install -qq protobuf-compiler python-pil python-lxml python-tk
    
    pip install -q Cython contextlib2 pillow lxml matplotlib
    
    pip install -q pycocotools
    
    cd /content/models/research
    protoc object_detection/protos/*.proto --python_out=.
    
    import os
    import sys
    os.environ['PYTHONPATH'] += ':/content/models/research/:/content/models/research/slim/'
    sys.path.append("/content/models/research/slim/")
    
    python object_detection/builders/model_builder_test.py

• Then follow the steps from this famous blog of Gilbert Tanner

Note: You might get errors when running the code block as above, or while running the python file as below. Keep fixing every errors, until TF is properly installed.

python object_detection/builders/model_builder_tf1_test.py

Tensorflow installed and I did inference as well.