Block identifier model trainer

This program is responsible for training a classifier for lego technic parts. Under the hood it uses mobilenet to extract features from the provided image and the k-nearest neighbor algorithm to find the type of block the provided image is closest to. The source code for this project, along with the dataset is available on github under GPL v3. Press the button below to start training the model:

Please note that the training process is mostly network bound. The learning for a particular example happens after an example has loaded, so you will not experience high cpu usage.