The company blog for Halide, an iPhone camera app, on iPhone camera hardware, RAW capture, computational photography, and app development.
HN commenters praise the ProRAW post for its technical detail, and cite its iPhone XS camera article to debunk a claim about phone cameras.
Understanding ProRaw[1][2] was a pretty great read on some of the technical capabilities.
It's hugely appealing to have the data accessible to humans. Google has amazing computational photography, but the human is out of the loop. You can shoot raw, & it gives you a post-HDR "raw". Apple gives you the raw image, and tells you what tone mapping they are suggesting via a Profile Gain Table Map. This is a huge huge huge boost to human agency, jump starts a relationship with technology, rather than just some blackbox magic being handed to the consumer. Apple is organizing & making accessible the information of computational photography.
Apple again makes the computational photography information available by storing a semantic map of the photo in ProRAW, identifying different subjects of the photo.
It makes me so happy to see someone supporting & enhancing the creative arts with technology, generating more information on what is happening within the system & allowing humans to participate in these expert systems. I hope we see more/some competition to open up computing to humans.
[1] https://news.ycombinator.com/item?id=25433047
[2] https://blog.halide.cam/understanding-proraw-4eed556d4c54
Videos can use inter-frame information to help infer sub-pixel details.
This post about "Super Resolution" is interesting because it starts with RAW format (which contains information about camera sensor arrangements), hence, the machine-learned model should not only memorize artificial details (what hair should be look like, what a tree-leaf should be look like etc, and use that to "hallucinate" a higher-resolution details, I liked to call this "hallucination" for that reason), but also relationship between complex interference of different sensors in their corresponding arrangements.
You can read more about RAW format and why exposing RAW format for photography is exciting (on everyday's camera, i.e. your phone) from this post: https://blog.halide.cam/understanding-proraw-4eed556d4c54
It's not Swift. App sizes are mostly increased by graphic assets, and libraries. Here is a 15,000 line swift app that compiles to 11 megs.
https://blog.halide.cam/one-weird-trick-to-lose-size-c0a4013de331
I'm skeptical Swift creates significantly larger binaries, let alone 3x-4x, but I'd still use it if so. Using optionals and stronger typing is makes it easier to write correct code, and anyway we can save users from wild pointer crashes is super valuable. Far more valuable than the cost of a couple megabytes of code in an ocean of PNGs and Cocoapods.
There were a few articles with actual content on this topic covered on Daring Fireball in recent months. Not sure what this blog blurb is adding to the conversation.
1. https://sensortower.com/blog/ios-app-size-growth
2. http://blog.timac.org/?p=1707
3. https://blog.halide.cam/one-weird-trick-to-lose-size-c0a4013de331
While I'm sure there are phone makers doing that, the one you're linking appears to have been debunked a while ago. [1]
[1] https://blog.halide.cam/iphone-xs-why-its-a-whole-new-camera-ddf9780d714c?gi=571ae9c7d382
Swift adds a few megs but you can minimize that impact by doing things like this guy did:
https://blog.halide.cam/one-weird-trick-to-lose-size-c0a4013de331