1 Install tensforflow

2 Install bazel

3 Set environment variables

IMAGE_SIZE=224
ARCHITECTURE="mobilenet_0.50_${IMAGE_SIZE}"

4 source tensor flow

source mytensforflowdirectory/bin/activate

start tensorboard in a separate terminalzzzz

tensorboard --logdir tf_files/training_summaries &

5 run use tensorflow_for_poets retrain script / example

python -m scripts.retrain --image_dir ./training_data/ --how_many_training_steps 500 --model_dir=./tf_files/models --architecture="${ARCHITECTURE}" --output_graph=tf_files/retrained_graph.pb --output_labels=tf_files/retrained_labels.txt

6 apply label

python -m scripts.label_image --graph=tf_files/retrained_graph.pb --image=testing_data/fairway8.jpg

Note:

Had to use converseen to convert png to jpg

Online SVM requirements:

• scikit-learn
• matconvnet
• matlab python hooks