Retrain a pre-trained Neural Network to recognize Images
- Python 3.6.2 (https://www.python.org/downloads/release/python-362/)
- Numpy (https://pypi.org/project/numpy/)
- Tensorflow (https://pypi.org/project/tensorflow/)
- Keras (https://pypi.org/project/Keras/)
- Create a folder in desktop, and extract all files to that folder. Create a folder tf_files inside the main folder. Place all images in
tf_files/[any_name]/folders.
The folders must be name of the objects inside. For eg. Folder names Cat will have images of cats
- Go to main folder and Run the following code in CMD:
python -m scripts.retrain --image_dir=tf_files/[any_name]/ --how_many_training_steps=1000 --model_dir=tf_file/model/ --input_height=299 --input_width=299
-
Store the Image to Classify in the main folder
-
In the main folder run the following in CMD:
python -m scripts.label_image --input_height=299 --input_width=299 --image=[file_to_classify].jpg
For more info on training an Image Classifier: https://codelabs.developers.google.com/codelabs/tensorflow-for-poets/#0
For doubts email me at: atinsaki@gmail.com