An image classifier built using TFLearn.
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.


An image classifier built using TFLearn.

Training A Model

import glob
images = glob.glob('*.png')
c = ImageClassify(['yes', 'not'], image_size=100, learning_rate=0.001)

Training a model is very easy. Simply feed in the list of classes you want to train, and set the image size and learning rate parameters. The prepare data function takes in a list of images. Images are expected to be in the following format: "class.number.jpg", by passing in the images like this we can label them without the need for another labels file. Calling the train_model method will train a model and save it with the name you passed in.

Making A Prediction

c = ImageClassify(['yes', 'not'], image_size=100, learning_rate=0.001)
results = c.predict_image('road_sign.jpg')

Making a prediction is also very easy. Initialize the class with the same parameters as before. Load your already trained model and then call the predict_image method. The function returns the predicted class name and the raw result.