Application of Convolutional Neural Network in Remote Sensing Image Recognition
- Convert the Texas Image into PNG format and split it into small tiles
2.Start Docker with local files available % docker run -it -v $HOME/tf_files:/tf_files gcr.io/tensorflow/tensorflow:latest-devel # my file: /tf_filesclassification/evergreen forest, shrub_scrub, cultivated_crops
3.Retrieving the training code cd /tensorflow git pull
4.Start your image retraining
python tensorflow/examples/image_retraining/retrain.py \ # retrain contains the code used to train image classification
--bottleneck_dir=/tf_files/bottlenecks
--how_many_training_steps 500
--model_dir=/tf_files/inception
--output_graph=/tf_files/retrained_graph.pb
--output_labels=/tf_files/retrained_labels.txt
--image_dir /tf_files/classification
5.use the label_image.py to load my graph file and predicts with it. I put label_image.py in the folder of ty_files.
python /tf_files/label_image.py /tf_files/testpicture/*picture name