Subscribe
Sign in
Start new thread
All
Fritz
p/fritz
Teach your mobile app to see
Visit Product
Dan Abdinoor

Alchemy by Fritz — The easiest way to convert a neural network to Core ML

Featured
4
•

Alchemy is a collection of utilities and tools for the mobile machine learning community. At times, deep-learning and AI can seem like a magical process of transformation.

The Grader tool is available today, and we think this is the easiest, fastest, and most comprehensive way to benchmark and convert ML models for mobile.

Replies
Best
Naomi Assaraf 🔥
I had to spend at least 10m figuring this out, so I’m just going to make it easier for everyone else who’s not in the #ML Community and wants to understand what’s being released: **You can’t build your machine learning algorithms on mobile. So, this nifty tool converts it for you to a mobile friendly file that can easily be used in your mobile app. It also tells you what your runtime is... if it’s too long, you need to change your algorithm so that it can calculate a faster result for your users. It also helps eliminate bias in images, so there’s pros to working with this product vs. Apple.**
Dan Abdinoor
Hunter
@nassaraf Great description! We should have made the description less technical and opaque. Hopefully the blog article we have linked to helps to explain more, including addressing the question of Apple vs. Alchemy.
Dan Abdinoor
Hunter
Hey Hunters! We built Alchemy to be a collection of utilities and tools for the mobile machine learning community. Everything is free and doesn’t require an account. We just updated our Grader tool to support benchmarking and conversion. We are initially supporting models from Keras (.h5 files) and converting to Core ML (.mlmodel files), and plan to add support for additional formats soon. Leave a comment with your favorite framework and we’ll try to make it happen! To benchmark or convert your Keras model, upload it to Grader. It will give you information about the inputs, outputs, compatibility, and runtime performance. Click the “Convert” button on the report screen and we’ll help you transform it into a mobile friendly format. Then you can download a file you can drop directly into your app in XCode. Read more about how we do it: https://heartbeat.fritz.ai/annou... If you’re interested in learning more about putting machine learning and AI into mobile apps, check out our publication Heartbeat (heartbeat.fritz.ai) and sign up for our newsletter at bit.ly/joinheartbeat. Thanks! Dan & Fritz Team
Damjanski
<3 <3 <3