360|iDev 2017 Conference Highlights
270 attendees and 59 speakers recently descended upon Denver, Colorado to take part in the annual 360iDev conference. 360iDev 2017 had much to choose from; with seven workshops on Sunday and a whopping 58 sessions over the next three days to choose from, it was easy to find find a mix of sessions tailored to […] By Tim Mitra.
Deep Learning on iOS – Shuichi Tsutsumi
In his talk, Shuichi Tsutsumi presented some interesting examples of machine learning in action. He explored the evolution and use of deep learning on actual iOS devices. He pointed out some shortfalls of trying to reach a cloud based service without a network signal. Shuichi covers the current state of MPSCNNs and BNNs, which have been available since iOS 10.
If you’re curious about the steps required to use deep learning in your apps, he covers these in his talk, including the steps for training a model, implementing a neural network, and implementing an interface. CoreML, he goes on to explain, also employs MPSCNNs and BNNs under the hood. The former is more GPU-efficient, while the latter is more CPU-efficient.
In his final demo, he shows how simple it can be to use CoreML. Choose a model; if necessary, convert it to a CoreML model; then drop the model into your project. Add the Vision framework and you’re off and running with an app that can (mostly) identify the objects around you.
TensorFlow on iOS – Taylan Pince
Taylan Pince starts off by saying that he should have titled his talk “What the heck is a Neural Network anyway?” He tells of his numerous years working on a client project, Field Guide: a collector’s guide of natural history. Having started around 2014, his team started out by exploring computer vision to categorize around 100 species. Using clever image processing techniques, the catalog grew substantially. Eventually, they moved to ImageNet, a research database with 14 million pre-trained images. Taylan then explained how images are weighted for predictability in a neural network.
In the last part of his talk, Taylan explores and compares Tensor Flow, CoreML, Metal Performance Shaders and the Accelerate framework. If you’re looking for a exploration of machine learning and CoreML, this talk, along with Shuichi Tsutsumi’s, are definitely worth checking out.
Playing Nice with Design – Ellen Shapiro
As developers, we can gain a lot by deploying effective and practical communication with our design team. The first challenge, though, is establishing a common terminology. Designers, iOS developers and Android developers can use terms that have very different meanings in their respective ecosystems.
To start, create a table to map out the common terminologies — and while you’re at it, do the same for font styles and colors used in your design language. The team Ellen worked on created an open source app, True Colors, to see the colors used on various devices. She also likes Sourcery to generate code to store values then share those values in files added to projects.
Creating a custom framework is another way to create building blocks that can be used in your apps.
Ellen also takes a look at the benefits and pitfalls of using your own frameworks. Designers can start their work on iPads and Playground Books to create resource files that the developers can run under Swift Playgrounds on iOS.
In summary, start with a small goal and build up with common paradigms; text, fonts, labels, margins, etc. This talk was full of tips that can be used to add intelligence and tools to your team’s communication.
Bonus: Check out Chris Wagner’s tutorial on Sourcery to see how to create useful templates for your team.
iOS with Continuous Delivery – Cassie Shum
Cassie explains the nuts and bolts of continuous delivery while covering a number of tools and workflow enhancements in this detail-filled talk. The difference between continuous deployment and continuous delivery is that while deployment to production is optional under continuous deployment, continuous delivery can and does deploy to deploy to production.
Continuous delivery, she says, has a reduced risk, as fewer lines of code can be delivered more frequently while changes are fresher in the developer’s mind.
She went on to break down the tools by phases: Build, Deploy, Test, and Release. Best practices include using clean architectures and design patterns to avoid the bloat of the “massive view controller”. Tools such as SwiftLint, ocLint and static code analysis make for better and consistent code. Deployment tools like fastlane automate the pain points, and HockeyApp and TestFlight get builds onto devices for testing.
This talk is packed with workflow enhancements and tools. Definitely check this one out.
Bonus: Check out Lyndsey Scott’s tutorials on fastlane to learn how to automate the drudgery of app deployment.
Fun & Games
The conference was more than just talks. There was some fun & games too!
Stump 360 Episode IV: A New Hope – Hosted by Tom Harrington
Presenting the…Experts?
The fourth annual “Stump 360” picked up where the WWDC favorite “Stump the Experts” left off. A rag-tag collection of “experts” took on the gathered audience in a game-show style battle of inane Apple trivia. The hosts presented questions to challenge the audience, who in turn wrote trivia questions on 3×5 index cards.
The event was rife with comedic moments, and most often useless trivia, with points awarded to each side. Prizes consist of extremely valuable 5-1/4-inch floppies that may have been overwritten, old eWorld and Newton stickers, and a vintage case for a PowerBook Duo battery — batteries not included. This session is a true highlight, and I look forward to many more years of Revenge of the Stump 360, or whatever they choose to call it.
Full disclosure: We did manage to stump some of the audience. However the score was close, as we “experts” were defeated by the audience members! :]
Game Dev Jam – Hosted by Ryan Polos
Every year I’ve attended 360iDev, there’s been an all-nighter dev jam where bleary-eyed developers show off their work first thing in the morning to the collected masses. This year, there were two apps employing ARKit and one watchOS app. The first was a game where players could shoot down pesky Tie Fighters. A second game placed a shuffleboard on a nearby surface then allowed players to send virtual rocks down the board. The watch app enabled wearers to watch and bid on eBay auctions.
The game dev jam and accompanying board game night provided a great way to socialize and collaborate with other developers from around the world.
Bonus: Subscribers can check out our screencasts on ARKit to see how easy it is to get started with ARKit.
We also cover ARKit in iOS 11 By Tutorials, which is available on our store.