Monthly Computer Vision Roundup #2 2017
Welcome back to our Computer Vision Roundups, our February Meetup was an almost-February-Meetup, taking place on March 1st. Good thing all of you got the memo – we were delighted to have started the year with so many more members in our Computer Vision Group on Meetup.com and more new faces at our regular meetups!
This time, our CTO and host of the meetup Daniel Albertini was focusing on a very hands-on topic: A tutorial in Deep Learning in iOS + Live Demo.
Deep Learning in iOS
This talk was catered to beginners in development with deep learning and showcased how a first project could be set up in iOS. Also, different options and frameworks were introduced.
Short Outline of the Talk:
- About iOS Developdas ment
- BNNS Functions
- About Tensorflow
- Tensorflow Deep MNIST Tutorial
After a quick intro about possibilities in iOS Development, we dove right into the Accelerate.framework – a framework from Apple, that helps in running very big computational and mathematical calculations directly on your phone. Therefore the framework is perfectly optimized for high performance on the CPU.
Within the Accelerate.Framework we find the BNNS Functions. In order to set up a neural network within iOS applications, the Basic Neural Network Routines (BNNS) are the way to go, as they provide highly performant inference for neural networks. You need to have already trained the data though, which can then be fed to the BNNS.
In order to train neural networks then, Daniel arrived at TensorFlow, an open source framework developed by Google for deep learning and machine learning research. In the Deep Learning on iOS Intro talk, Daniel uses TensorFlow for training a neural network with a training set of thousands of digits – the MNIST database of handwritten digits.
In order to re-enact the Live Demo, please do follow the Deep MNIST for Experts Tutorial by TensorFlow! This tutorial guides you through training a model with TensorFlow in order to detect digits in the MNIST data set, get data about the accuracy of your model and concludes with the training of neural networks in order to improve performance and accuracy.
A quick info for all of you who is interested in TensorFlow: In February the first TensorFlow Dev Summit was held in Mountain View by Google Developers – the videos are now online on their Youtube Channel!
More Deep Learning Talks
For a Deep Learning Introduction in the browser with ConvNetJS and CaffeeJS, have look at the presentation from last July’s Meetup by Christoph Körner!
Hold a talk yourself!
You have a project or topic you would like to talk about or you know someone, who would like to share his/her experiences and knowledge at our Computer Vision Meetup? Please contact us!
It is great to see how our community is growing each month, so if you haven’t already – don’t forget to join our meetup group ! ;)
QUESTIONS? LET US KNOW!