Jump to content

zzy

Verified Members
  • Posts

    219
  • Joined

  • Last visited

Posts posted by zzy

  1. Hi @Tima

    I've spent some time to check if we can support x86 in our current code base. There are some blockers:

    1. There are currently some dependencies (internal and external) missing x86 support.
    2. Performance of x86 is worse than x64, we need some time for tuning before release. Please note even after tuning, x64 still runs faster than x86.

    Based on the current situation, I would expect at least 2-3 months before we can release the x86 version.

  2. Hi @Tima

    We only support x64 binaries in our SDK release for Windows due to the following reasons:

    1. Most VR developement tools supports x64 (Unity, Unreal) and are dropping x86 supports.
    2. VR applications usually can benefit from 64 bit advantages.
    3. Our QA process can be simplified.
  3. Hi @LeRuf

    After a few searching with Google, I think this might be a bug of Unity.

    Here are related issues and forum discussion:

    https://issuetracker.unity3d.com/issues/android-crash-with-gsl-mem-error-kgsl-sharedmem-alloc-ioctl-failed-on-adreno-gpu-devices-1

    https://issuetracker.unity3d.com/issues/android-crash-with-gsl-mem-error-kgsl-sharedmem-alloc-ioctl-failed-on-adreno-gpu-devices

    https://forum.unity.com/threads/fixed-memory-issues-on-android-since-2018-3-0f2.599551/

    Although the issues are claimed to be fixed, but people are still experiencing this bug. You can even find people taking about Focus Plus in the first link.

    Recommended solution:

    1. You can try the latest 2019.3 version (which should be 2019.3.13 according to this page: https://unity3d.com/cn/get-unity/download/archive)
    2. I would recommend using 2018.4 LTS version for stability.
    3. You to create a bug report to Unity to see if this can be fixed.
    • Like 1
  4. Hi @LeRuf

    From the log your pasted, I think the problem is here:

    Quote
    0001/01/01 00:00:00.000 -1 -1 Info  --------- beginning of crash
    2020/05/13 11:37:57.041 22605 22799 Fatal libc /buildbot/src/googleplex-android/ndk-release-r20/external/libcxx/../../external/libcxxabi/src/abort_message.cpp:73: abort_message: assertion "terminating with uncaught exception of type cv::Exception: OpenCV(3.4.7) D:\Project\deeplearning\aristo\3rdparty\download\opencv-3.4.7\modules\core\src\alloc.cpp:72: error: (-4:Insufficient memory) Failed to allocate 32768 bytes in function 'OutOfMemoryError'

    This seems like a out of memory. Some ways to fix:

    1. Build the app for arm64 using il2cpp. It should have more memory available than armv7.
    2. Remove some objects from the scene.
    3. I think some player settings can also help this, e.g. vertex compression, reduce number of shaders, etc.

     

  5. Hi @coloal

    HandModel scene does work on Android phone, since this scene requires skeleton mode which is not supported on Android phone now.

    I haven't used it with Vuforia or AR engine. Since we use NDK camera to get rear camera image, and Android only support one consumer for each camera, I think this will not work with other SDKs that also use rear camera.

  6. Hi @nocanwin

    Hand tracking does not support running with CPU on Windows, so this error is expected behaviour.

    As for the scene performance, do you have some break down about the time? What is the difference between 12ms and 5ms in Unity profiler? Can you identify some scripts that takes very long time, or the whole performance is slow? Please also check Windows Task Manager and see if CPU is at 100% utilization.

  7. Hi @nocanwin

    If the sample project runs at 9ms, then I would suggest you to check what is in your scene that makes the execution slow. I would suggest to use the same project for both sample and your scene to make a even compare.

    What is your game engine, Unity or Unreal? What's your HMD? Do you have some detailed profiling result (provided by Unity or Unreal) that can show which functions is taking a very long time?

  8. Hi @davide445

    Glad to know you have mounted leap motion successfully. About your question, wearing a glove does not help for Vive Hand Tracking SDK, since our SDK is trained with hands without gloves. Here are some suggestion you can try:

    1. Roll up sleeves and take off your watch if any
    2. Avoid light sources in the background, including window and lights
    3. Try to make your room brighter (e.g. turn on the lights even in day time), but don't look at light sources directly
  9. Hi @davide445

    I haven't used Leap Motion for about 3 years, so I'm not sure if there is any place feasible on Cosmos for it. I think you have to test it yourself. As for the USB port, you might need the to connect to all the way back to your PC (ugly, but should be working).

    Also, there is a new product that is announced recently: the Cosmos XR. We would expect hand tracking to perform better on it since camera is better.

  10. Hi @davide445

    Thanks for your video test. I think the problem is when your hand is small (e.g. fist gesture) or your hand is a little further from the camera, the result is not good enough. We will try to see how can we improve this situation.

    As for the v1 release, we currently don't have a solid date for v1, but we definitely hope to have it this year. That's all I can say about the release date right now. We will keep using the same license as the current release.

    Also, please note that due to the nature of the cameras, some self-occluded gestures may not have good performance, even in the v1 release. I would suggest you to avoid using such gestures in your application.

  11. Hi @davide445

    Please let me know what problem you encountered in UE4 so we help to solve the problem.

    Regarding to the video, I think this is what to except at the moment, basically all your hands can be tracked correctly with small jitterness. There are some occasional wrong track result (especially for left hand) and we are trying to solve that issue.

    To improve the situation, I think here are the things you can try:

    1. Try to increase the overall brightness of your environment. We have got reports that brighter room has better performance. Please avoid sunlight directly on your hand.
    2. Roll up your sleeves if any, this can help to reduce the chance of wrong track restuls.
    3. Try a cleaner environment.

    Please let me know if this helps.

  12. Hi @davide445

    Since all your actions are quite similar to natural input, I think hand tracking is easier to learn than controllers, especially for non-gamers.

    The SDK is in beta status because we think there are still much room for performance improvedment, but feel free to use it in your commercial projects.

    As for the support, forum is the only public support for hand tracking currently. But if you already have Vive commercial support, you can raise your questions there and they will loop us to answer the questions for you.

    @VibrantNebula, can you please help to comment about the commercial support?

    • Thanks 1
×
×
  • Create New...