Juxtaposer is an iOS app for easily combining multiple images into funny and/or artistic photomontages. The new Magic Crop feature just added in version 4 uses a Machine Learning model to let you cut out an element from an image in seconds.
Hi y'all,
I am an indie iOS developer and just added an (I like to think) very cool and useful Machine Learning based feature to my now 12 year old (launched in 2008!) photo compositing app Juxtaposer.
I have been working on this new feature for over a year, researching different deep neural network models, developing the model and training code, gathering and preparing training data, training the model with it, and then creating the iOS UI and now finally shipping it. I am really happy to get this into the hands of users now and I hope some of you in the Product Hunt community will find it interesting.
The ML model runs on the iOS device, which makes it privacy preserving (your photo data doesn't leave the device), and also very fast thanks to the Neural Engine in Apple's latest devices.
The image segmentation takes about 130 ms on an iPhone 11 Pro, and probably significantly less still on the new iPhone 12. I'll find out later today when my wife gets her 12 mini.
I've wanted to add some form of semi-automatic image segmentation to this app pretty much from the beginning, and now 12 years later I finally did 😀
I am happy to answer any questions and of course happy to get feedback.
Juxtaposer