nerf_tracking

start the virtual environment

cd ml source env/bin/activate

#building the model python -m scripts.retrain
--bottleneck_dir=tf_files/bottlenecks
--how_many_training_steps=4000
--model_dir=tf_files/models/
--summaries_dir=tf_files/training_summaries/"mobilenet_0.50_160"
--output_graph=tf_files/retrained_graph.pb
--output_labels=tf_files/retrained_labels.txt
--architecture="mobilenet_0.50_160"
--image_dir=tf_files/flower_photos
--random_brightness 15

running the test image script.

python -m scripts.label_image
--graph=tf_files/retrained_graph.pb
--image=tf_files/test_images/IMG_0267.jpg

compile that graph!

./bonnet_model_compiler.par
--frozen_graph_path=retrained_graph.pb
--output_graph_path=retrained_graph.binaryproto
--input_tensor_name=input
--output_tensor_names=final_result
--input_tensor_size=160

On Pi

./test_run_model_on_bonnet.py
--model_path ~/AIY-projects-python/src/aiy/vision/models/retrained_graph.binaryproto
--input_height 160
--input_width 160

Run the model!!

./mobilenet_based_classifier.py
--model_path ~/AIY-projects-python/src/aiy/vision/models/retrained_graph.binaryproto
--label_path ~/AIY-projects-python/src/aiy/vision/models/retrained_labels.txt
--input_height 160
--input_width 160
--input_layer input
--output_layer final_result
--threshold 0.8
--preview
--show_fps