This article and video tutorial will help you get up to speed with generating synthetic training images in Unity. You don't need any experience with Unity, but experience with Python and the fastai library/course is recommended. By the end of the tutorial, you will have trained an image segmentation network that can recognize different 3d solids.

Continue reading

Creating a Deep Learning iOS App with Keras and Tensorflow

Take the Food Classifier that we trained last time around and export and prepare it to be used in an iPhone app for real-time classification.

Continue reading

Secure 'Serverless' File Uploads with AWS Lambda, S3, and Zappa

Deploy a Python-based Zappa microservice onto AWS Lambda to facilitate direct browser to S3 file uploads. A React-based Fine Uploader component is hosted on S3. All traffic is encrypted (for free) with HTTPS/SSL!

Continue reading

Accelerating Deep Learning with Multiprocess Image Augmentation in Keras

By adding multiprocessing support to Keras ImageDataGenerator, benchmarking on a 6-core i7-6850K and 12GB TITAN X Pascal: 3.5x speedup of training with image augmentation on in memory datasets, 3.9x speedup of training with image augmentation on datasets streamed from disk.

Continue reading

Food Classification with Deep Learning in Keras / Tensorflow

Work with a moderately-sized dataset of ~100,000 images and train a Convolutional Neural Network to classify the images into one of 101 possible food classes. Side excursions into accelerating image augmentation with multiprocessing, as well as visualizing the performance of our classifier.

Continue reading