"The industry verdict on 3D sensing [inside smartphones] varies widely — from “why bother?” to “it’s the future.” While some interpret the lack of iPhone X competitors with full-blown 3D sensing technology as a lack of market interest, others disagree.
While Apple’s TrueDepth had by 2017 established the trend for 3D front-facing cameras, Yole acknowledged that the wave [for 3D sensing adoption] “has started on the conservative side in terms of volume.”
The 3D sensing argument took a positive turn when Chinese smartphone OEMs including Xioami, Oppo, and Vivo unveiled their plans for 3D sensing over the last few months. Although Cambou is sure about front-facing 3D, he remains skeptical of use cases for 3D in rear-facing cameras. Pointing out a lack of momentum for VR and AR, he explained that neither the augmented reality sales pitch nor augmented gaming are yet proven on the market.
Recent quarterly financial calls held by STMicroelectronics and ams revealed that “they are almost overly confident” that 3D sensing will go inside both the front and rear cameras of smartphones, observed Cambou.
So what percentage of smartphone cameras will have 3D cameras? What’s the penetration ratio? Yole predicts that a 1.4% penetration ratio in 2017 will grow to 55% in 2023.
Huawei... is making a big investment in digital photography, said Cambou. In comparing the size of active matrix (used for traditional photography), he calculated that Apple is devoting only 52 mm², while Samsung has committed to 91 mm² and Huawei 112 mm²."
Component suppliers need to remain up to speed with advances in CV. Advances in SfM could throw off cellphone camera module sales projections and turn stereo/structured/tof into a nice to have for higher end phones.
ReplyDeleteWe should start with abbreviations. CV is computer vision I guess. SfM has me stumped.
DeleteStucture from motion.
ReplyDeleteRequires (mostly) static scene + motion, handles just a subset of apps compared to a proper realtime 3D cam.
You assume 3D is all about static modeling.
ReplyDeleteWe see and work on many dynamic use cases which need good depth on every frame
I hope from phone manufacturer opens up API for developers to access 3D data -- better still, live stream data -- dynamic use case as you mentioned.
DeleteGreat to know mantis vision is being adopted already in mobile phones !
Computer vision and machine learning are definitely enhancing the capabilities of 2D cams, but instantaneous and accurate 3D sensing is needed for best user experience. The end consumer will unlikely wait for calibration or SfM-like processes to kick-in in their daily usage.
ReplyDelete