Re:Core ML vs. ML Kit: Which mobile machine learning platfor...
by mtgx on Friday July 06, 2018 @11:57AM
(#17472451)
> The on-device version of ML Kit offers low accuracy compared to the cloud version, but at the same time it offers more security to user data.
I understand why that could be true, but I also wonder if Google did that on purpose a little bit, for the same reason they make you configure their Wi-Fi routers and other devices over the internet.
Re:Core ML vs. ML Kit: Which mobile machine learning platfor...
by abdinoor on Friday July 06, 2018 @01:16PM
(#17473185)
(Full disclosure: I work at Fritz, and we publish Heartbeat).
There are a couple of legitimate reasons that the on-device model is lower accuracy: compute and size. Models are often reduced in complexity when running on edge devices, but cloud-based GPU (TPU in Google's case) models can require far more calculations in order to squeeze that bit of accuracy. As for model size, a 100MB model in the cloud is not a big deal, but having to download a 100MB model on every edge device is expensive in both time and bandwidth.
Outside of these reasons, Google/Firebase may want on-device capabilities limited in order push adoption of their cloud ML services.
Re:Core ML vs. ML Kit: Which mobile machine learning platfor...
by Razengan on Friday July 06, 2018 @08:01PM
(#17476026)
> compute and size ... having to download a 100MB model on every edge device is expensive in both time and bandwidth.
The WWDC 2018 ML sessions showed some solutions to address these issues. (I don't exactly recall which as I'm not using ML yet, so I skimmed through them.)
Re:Core ML vs. ML Kit: Which mobile machine learning platfor...
by jchw on Friday July 06, 2018 @12:49PM
(#17472916)
To be fair, Linksys also makes you configure their consumer routers over the internet, and I trust Linksys a lot less.
Footnote: You can do a "local" login on a Linksys smart router, but if your internet is not working during setup it will refuse to even initialize. It made me reset the NVRAM because my internet wasn't connected yet. If you do refuse to create an online account, the internet service is still enabled by default.
Re:Core ML vs. ML Kit: Which mobile machine learning platfor...
by upgoat on Saturday July 07, 2018 @03:47AM
(#17477569)
Haven’t used ML Kit but CoreML has been very simple to setup (few lines of python + drag n drop into Xcode) and very fast on device. You can do video stream classification live
Re:Core ML vs. ML Kit: Which mobile machine learning platfor...
by amelius on Saturday July 07, 2018 @04:53AM
(#17477727)
Yes, but the point is that if the open, portable solution gets a sufficiently large userbase, then the platform owners will be highly motivated to provide/implement efficient support for the open standard.
I understand why that could be true, but I also wonder if Google did that on purpose a little bit, for the same reason they make you configure their Wi-Fi routers and other devices over the internet.
There are a couple of legitimate reasons that the on-device model is lower accuracy: compute and size. Models are often reduced in complexity when running on edge devices, but cloud-based GPU (TPU in Google's case) models can require far more calculations in order to squeeze that bit of accuracy. As for model size, a 100MB model in the cloud is not a big deal, but having to download a 100MB model on every edge device is expensive in both time and bandwidth.
Outside of these reasons, Google/Firebase may want on-device capabilities limited in order push adoption of their cloud ML services.
The WWDC 2018 ML sessions showed some solutions to address these issues. (I don't exactly recall which as I'm not using ML yet, so I skimmed through them.)
Footnote: You can do a "local" login on a Linksys smart router, but if your internet is not working during setup it will refuse to even initialize. It made me reset the NVRAM because my internet wasn't connected yet. If you do refuse to create an online account, the internet service is still enabled by default.
DirectX vs. OpenGL for example, or Adobe RGB vs. DCI-P3 on recent wide-color devices, or the convenient but clunky abomination that is Electron.