5 ms·

Core ML vs. ML Kit: Which mobile machine learning platform do you prefer?

55 points by austin_kodra 8y ago

mtgx 8y ago
> 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.
- jchw 8y ago
  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.
- abdinoor 8y ago
  (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.
  - Razengan 8y ago
    > 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.)
gok 8y ago
Turi Create is a nice alternative to Create ML for building Core ML models.
amelius 8y ago
I'm hoping developers choose the long-term sensible option, which is the multi-platform one.
- Razengan 8y ago
  Multi-platform solutions don't always seem to be the most optimal/efficient solution on each specific platform. 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.
  - amelius 8y ago
    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.
upgoat 8y ago
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

55 points by austin_kodra 8y ago

mtgx 8y ago

> 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.

jchw 8y ago
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.
abdinoor 8y ago
(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.
- Razengan 8y ago
  > 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.)

gok 8y ago

Turi Create is a nice alternative to Create ML for building Core ML models.

amelius 8y ago

I'm hoping developers choose the long-term sensible option, which is the multi-platform one.

Razengan 8y ago
Multi-platform solutions don't always seem to be the most optimal/efficient solution on each specific platform. 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.
- amelius 8y ago
  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.

upgoat 8y ago

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