If you've been keeping up with Honor rumors lately, then you probably heard the company is planing a successor, or update, of sorts to its quite successful View 10. Apparently called the Honor 10, the device is expected to rock the same Kirin 970 flagship chipset, like its sibling, borrow the Huawei P20's 12MP, plus 20MP dual camera setup and make its first official appearance on April 19.
According to some recently unearthed documents, allegedly pertaining to the Honor 10, its headlining feature might be something called "semantic image segmentation mode". What it is, in practise, is a system to achieve multi-scene detection within a single photo. The Kirin 970 already has some advanced AI camera features that allow it to recognise things like pets, snow, beaches, mountains, etc., on the device itself, thanks to the powerful on-chip NPU. This new algorithm, apparently, extend that functionality even further, allowing the camera to recognise more than one feature within a still, precisely determine their position and apply localized effects and tweaks as needed.
Unfortunately, the leaked document is not available in full, but the source did provide a couple of sample photos, allegedly taken on the upcoming Honor 10, with the mode turned on and off. Also, a few other thumbnail, showcasing how the detection actually works in some scenes. It looks quite reminiscent of Google's own DeepLab-v3+ algorithms for image segmentation. But unlike those, Honor's solution runs offline, on the device itself.
It is really early to say whether this information is entirely true and if so, how good the effects actually are. On paper, the whole things sounds very promising. Plus, given that, as far as we know, it is tied to Huawei's Kirin 970 and its NPU, there is the distinct potential for the algorithm to make its way to other devices, like the P20 duo or the Mate 10 duo.
As for the Honor 10 itself, it is said to have a 5.8-inch FullView display, up to 6GB of RAM and 256GB of storage. Color-wise, leaks seem to point towards a hue finish on the back, that may change according to surface temperature - reddish where the hand goes and bluish elsewhere.