Giter Site home page Giter Site logo

ju1ce / mediapipe-vr-fullbody-tracking Goto Github PK

View Code? Open in Web Editor NEW
394.0 19.0 39.0 166 KB

A repository using the MediaPipe API for fullbody tracking in VR with a single camera.

License: MIT License

Batchfile 0.59% Python 94.37% HTML 5.04%
fbt fullbodytracking vr vrchat osc

mediapipe-vr-fullbody-tracking's Introduction

Mediapipe-VR-Fullbody-Tracking

A test repository using Mediapipe for fullbody tracking in VR with a single camera.

This is still a work in progress, but an executable is now available for anyone to try out. Compared to my other free fbt project, ApriltagTrackers, this works less acurately, you need far more room and depth detection is not the greatest, but it has the benefit of not needing trackers and being far easier to setup and use!


For any questions or bug reports, please write into the ApriltagTrackers discord!


Using VRChat OSC:

The newest beta release supports VRChats OSC trackers. It should be possible to use it for FBT on standalone quest VRChat. To download it, go to the releases tab and download the 0.7 beta version.

To use OSC, change the backend parameter to VRChat OSC. If you will run VR on the same PC as mediapipepose, or you will use Quest 2 through link/VD/ALVR you can leave the IP at default. If you use quest standalone, enter the local IP address of your quest. If you use standalone, make sure to enable the WebUI option as well.

The process of using the OSC version is the same as regular, with a few differences:

  • When doing automatic calibration, it will not adjust Y rotation. Adjust it manualy until the skeleton aligns with your body.
  • Smoothing window is not supported, adjust smoothing using the Additional Smoothing parameter instead. It works better anyway

The WebUI:

If you are on Quest standalone, you cannot access the normal UI while wearing the headset, making calibration that way quite difficult. Thats where you enable the webui.

When starting up detection, the console will log the IP and port that you need to connect to. Open the web browser on Quest, enter the IP and port (NOT the 127.0.0.1 one) into the address bar, and open the page. A minimal UI should open, giving you control over the calibration parameters.

NOTE: This build is still in beta, so make sure to report any issues either in the github issues, or on the discord.

How to run:

An executable version has now been added to make running this easier. Download it from here. Unzip the folder anywhere you want.

First, install the driver by running the install_driver.exe from /driver_files.

Before you start the program, put the camera you will be using into a good spot. The camera HAS to see your entire body, and it helps if it is above you tilted downward!

Then, you should be able to run the program with start_mediapipepose.bat file. A setting window should appear with the initial settings:

  • IP or ID of camera: If using IP Webcam on android, this will be an IP address. Enter it in the form http://:8080/video. If using a wired webcam, this will be a number. Usualy 0, but sometimes 1 or 2 etc... If 0 doesnt work, try others.

  • Camera width and height: The resolution that the camera will be opened with. Higher values do not mean better tracking, so you should leave this at default unless you have issues with the camera, such as not opening or being really zoomed in.

  • Attempt to open camera settings: Try to open an additional window with settings for the camera, such as gain, exposure...

  • Enable experimental foot rotation: Foot rotation! May not work at all angles, however, so you may want to disable this.

  • Disable hip tracker: Disable hip tracking, if you wish to use owotrack for hips instead.

  • Enable advanced mode: Advanced mode enables you to edit more settings. Only enable this after getting used to the software, as the additional settings can break detection if used incorrectly!

  • Save and continue: Save the settings and start detection!

After pressing the save and continue button, a window with the camera feed and a window with runtime settings will appear! If they do not, check troubleshooting steps below.

Now, put on your headset and go into vr, but dont launch vrhcat yet: we still need to calibrate the tracking.

To do so, open your SteamVR dashboard (from inside vr), navigate to your desktop, and to the settings window. To calibrate, stand straight, look straight ahead, and press the Recalibrate button. Your head must look straight forward from your body, or rotation will be calibrated wrong! Now, if you look down, you should see three vive trackers that should follow your body, and you should be ready to play!

The other settings on the window:

  • Pause/Unpause tracking: Pause detection, if you want to freeze your fbt in place when sitting down.
  • Smoothing window: The smoothing window in seconds. The higher it is, the smoother will motion be, but with more latency and more "wobblyness". Good values are 0.3 for speed, 0.5 for balanced and 0.7 for smoothness. You may also want to disable it and only use the next option.
  • Additional smoothing: A diffrent kind of smoothing, from 0 to 1. Enabling this will reduce jitter and wobblyness by a lot, but also increases latency. A good value is 0.5 for speed, 0.7 for balance, and 0.9 for smoothness.
  • Camera latency: The latency of your camera. Increase to reduce delay, but at the cost of more shakyness. Note that it will not work when smoothing window is disabled. Should probably never be larger than 0.1.
  • Image rotation: Rotate the image to ensure it is upright, and mirror it if the camera feed is mirrored.
  • Log frametimes to console: Dump the detection speed to console, for debug reasons.

The window also contains sliders for manual calibration of values. If automatic calibration doesnt allign the skeleton well enough, you can refine it manualy using the sliders and buttons.

Your controllers may dissapear or no longer work when you use it the first time. If that happens, restart steamvr.

Automatic scale calibration may not work for some people, mostly on lighthouse tracked headsets. If you notice your feet are far above the ground, disable automatic scale calibration and use the slider instead.

Advanced settings

NOTE: Advanced settings will give you access to all settings, but plenty of them are still best at default, and may break detection otherwise! Read what they do below carefully!

When you get used to the software a bit and get it working well on normal mode, you can enable advanced mode to get access to a few new options, to further finetune it. It adds options to both the init settings and runtime settings:

New init settings:

  • Maximum image size: The image size is limited to this value, even if the cameras resolution is greater, to reduce processing costs. Mediapipe runs on very low res anyway, so no need to change this. Will also make the camera window larger.

Some DEV options, that you probably dont want to use:

  • Preview skeleton in VR: Instead of spawning just three trackers for legs, the entire skeleton will be spawned and shown 2 meters in front of you. Useful for visualizing how well it works. Dont use for actual gameplay!
  • Spawn trackers for hands: Try to emulate controllers by spawning trackers for your hands. Works pretty bad, so I would recommend against using this.

MediaPipe estimator parameters:

  • The only one you really want to change here is Model complexity. If you have a very slow PC, you can set it to 0 to speed it up a bit at a reduction of accuracy. On the other side, if you have a beast CPU, you can try to set it to 2 for increased accuracy. 1 seems to be the best balance, however, so you probably want to stick to that.
  • Other parameters are described here: https://google.github.io/mediapipe/solutions/pose.html

New runtime settings:

  • The smoothing parameters now have two values, that you can switch between. Useful to if you want to have two profiles, one for sitting that has more agressive smoothing, and one for movement, that has less smoothing.
  • HMD to neck offset: This is the offset from the hmd to the base of your neck, that is necessary to align the mediapipe skeleton to the hmd. First value is left/right, second is up/down, third is forward/back. Default is 0.2 meters down, 0.1 meter back and should work well enough. If tilting your head side to side or looking up and down moves your legs, it is set incorrectly. Check Troubleshooting to fix it.

Troubleshooting

The app crashes with an ImportError: DLL load failed:

This seems to be a problem with some windows "N" versions, which seem to miss some media features: Installing the media feature pack should fix it. Read more about it here: https://stackoverflow.com/a/54321350

The camera doesn't open:

For USB webcams:

  • Try a few diffrent IDs, 0, 1, 2 etc... If you have many virtual cameras installed, the number can go up to 10!
  • Check that the camera works in other apps, such as Discord
  • Check that the camera isnt opened in any other apps while trying to use it for MediaPipePose
  • Set the camera width and height to the resolution of the camera
  • Plug the camera into another USB port
  • Check/uncheck Attempt to open camera settings and try all the IDs again

For IP Webcam:

  • Ensure that the stream works when accessing it from the browser (just input the IP into the adress field of a web browser, it should connect to the stream)
  • Ensure that your phone is using WiFi and is on the same network as the PC. (the pc can be connected through ethernet, as long as its to the same router)
  • Check for typos. Its really easy to miss a number or a letter somewhere.

Connection to SteamVR fails after 10 retries:

  • Make sure SteamVR is open and a headset is connected
  • Make sure that you installed the driver with /driver_files/install_driver.exe, and that the driver installed properly
  • Open SteamVR settings -> startup -> addons, and make sure that apriltagtrackers is enabled
  • If SteamVR runs in administrator mode, you may have to run mediapipepose by running /bin/mediapipepose.exe as administrator
  • If using for VRChat, you can also just use OSC instead

The tracking is unuseably bad:

While the tracking is not perfect even under ideal conditions due to no depth information, there is still a few things you can do to improve it:

  • Make sure that your entire body is seen on the camera, including your head!
  • Make sure there is enough light in the room. and there shouldnt be any light sources behind you, such as any bright windows. Try to get as much lighting as you can from the direction of the camera.
  • Mind what you wear: you need to be in contrast with the background, so dont wear clothes the same color as the background and avoid clothes with patterns.
  • Play with the smoothing and latency parameters a bit. To learn what exactly they do, we have some nice graphs here.

Looking up/down moves my avatar up/down, or tilting head left/right moves my avatar left/right:

To fix this, you will have to finetune HMD to neck offset. It is only available when advanced mode is on, and is on the runtime GUI. The first value is x position(left/right, you want to keep this at 0), second is y (up/down) and third is z(forward/backwars). To correct this, follow the steps below:

Tilt your head to the left, then to the right, and observe what happens to the avatar. If your avatar moves with you, the y value is too high, and you want to reduce it (from -0.2 to -0.22 or so). If the avatar moves in the oposite direction from you it is too low, and you want to increase it (from -0.2 to -0.18 or so). Do this in small steps until your body no longer moves.

Now look up/down. If the avatar moves with you, the z value is too low, and you want to increase it (from 0.1 to 0.12 or so). If it moves the oposite, the z value is too high, and you want to decrease it (from 0.1 to 0.08 or so). Again, repeat in small steps until your avatar no longer moves with you!

All the trackers are 2 meters away from me:

The DEV: Preview skeleton in VR option will do this in order to preview the tracking, sort of like a mirror. This makes the tracking unusable for games though, so disable this option. And restart SteamVR after doing so.

The trackers are way too high:

If the trackers are far off the ground, or even above you, height calibration seems to have failed. Try to disable automatic scale calibration. This should enable a new slider to calibrate your height manualy.

Trackers dont allign with my body:

For automatic calibration, make sure that you stand straight and look straight ahead or some parts may not be calibrated well enough. If automatic calibration just doesnt seem to work, you can also use the sliders to calibrate it manualy. Rotation y will rotate your body around the up axis, rotation x will tilt you forward/backward, and rotation z will tilt you left/right. Adjust the values until the trackers align with your body!

Trackers don't show up in VR, status window says "Move to wake up trackers"

Problems with trackers may appear when using Additional Smoothing. Try disabling it.

This problem should now be fixed. If it still occurs, please let us know either by opening an issue or through the discord!

In vrchat, both legs stick together

First, make sure that the trackers align with the avatar properly. If they don't, you may need to recalibrate. On some avatars, you may also have to stand with your legs further apart to prevent this.

In vrchat, trackers dont work and cause issues

If there are any weird issues inside vrchat and nothing else works, it is usualy caused by mods. Mods can affect full body tracking systems in weird ways, even if the mod doesnt have anything to do with tracking. Try to uninstall all mods, even if they seem irrelevant.

Alternatively, you can also use the --no-mods vrchat launch option.

Other tips:

  • A wider FOV can help you have more room to move around, and ensure you can stand closer to the camera. If your camera/phone supports that, make sure to use it!

MediaPipe version

Thanks to John_ on the ApriltagTrackers discord, who reminded me that MediaPipe pose does in fact have 3d positions as well, the script was modified to use that version. It seems to be faster, pretty accurate, and should be easier to setup. Beside not needing the 3d model, the instructions on how to use are the same.

The movenet version has now been removed from the repo.

Information for devs:

Any needed libraries are listed in install_libraries.bat. The mediapipepose.bat should run through a local python installation, but it seems broken right now, so just use a global installation of python 3.9.

mediapipe-vr-fullbody-tracking's People

Contributors

byronbytes avatar ju1ce avatar realdragonturtle avatar yoyobuae avatar

Stargazers

 avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar

Watchers

 avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar

mediapipe-vr-fullbody-tracking's Issues

issue with conecting my phone camera

Importing libraries....
Reading parameters...
Save file not found, will be created after you exit the program.
Opening camera...
Connecting to SteamVR
Starting pose detector...
INFO: Created TensorFlow Lite XNNPACK delegate for CPU.
[ERROR:[email protected]] global D:\a\opencv-python\opencv-python\opencv\modules\videoio\src\cap.cpp (166) cv::VideoCapture::open VIDEOIO(CV_IMAGES): raised OpenCV exception:

OpenCV(4.5.5) D:\a\opencv-python\opencv-python\opencv\modules\videoio\src\cap_images.cpp:253: error: (-5:Bad argument) CAP_IMAGES: can't find starting number (in the name of file): http://192.168.0.101:8080/video in function 'cv::icvExtractPattern'

Camera opened!
Exception in thread Thread-2 (camera_thread_fun):
Traceback (most recent call last):
File "threading.py", line 1009, in _bootstrap_inner
File "threading.py", line 946, in run
File "mediapipepose.py", line 67, in camera_thread_fun
AssertionError: Camera capture failed! Check the cameraid parameter.
Traceback (most recent call last):
File "mediapipepose.py", line 185, in
AttributeError: 'NoneType' object has no attribute 'copy'
[12204] Failed to execute script 'mediapipepose' due to unhandled exception!
Press any key to continue . . .

Is it possible without SteamVR? And Multi-People

It's great application!!
I am a Game developer. I like this soooo much!
Q1. Is it possible to share the tracking result to other APP without VR? Maybe OSC to C# or Unity?

Q2.
I've tested 2 people, it's so stable. But will only tracking the last body only.
is it possible to track more than one people?

Thank You.

Linux Installer gets vr path wrong with steam flatpak

Normally the installer searches these two paths:

  • ~/.local/share/OpenSteam/steamapps/common/SteamVR
  • ~/.local/share/Steam/steamapps/common/SteamVR/bin/vrpathreg.sh

On the steam flatpak, the vr path is here: ~/.local/share/OpenSteam/steamapps/common/SteamVR
Manually editing the vrpath variable in install_driver.sh and uninstall_driver.sh lets it work correctly.

could not find config file. ( C : /Program files (x86)/Steam/config/steamvr . vrsettings)

that is what it says when I try to install_driver. I tried looking under C : /Program files (x86)/Steam/config, but the closest thing I could find to steamvr was steamapps. I couldn't open it as a file it wouldn't give me the option. Then I tried just opening it, but it only started up the Oculus and steamvr apps. I don't know what to do. Please help.

mediapipe pose vr not working

so basically, when i try to connect it to steam vr it gives my this bull crap: Importing libraries....
Reading parameters...
Save file not found, will be created after you exit the program.
Opening camera...
Connecting to SteamVR
Starting pose detector...
WARNING: Logging before InitGoogleLogging() is written to STDERR
F20221104 12:44:52.123401 23892 collection.h:55] Failed to get tag "C3__PACKET" index 0
INFO: Created TensorFlow Lite XNNPACK delegate for CPU.
*** Check failure stack trace: ***

Mediapipepose error

I tried beta 0,7 and 0.6 version.
I installed and uninstalled the drivers.
I checked addons in steamvr.
I tried cam id's between 0-2 some 0 and 1 had no problem.
I only tried with steamvr but it doesn't matter bec in 0.6 it doesn't ask which app
How can I solve this?
Mediapipe Pose Error

Traceback (most recent call last):
File "mediapipepose.py", line 181, in
File "mediapipepose.py", line 60, in main
File "mediapipe\python\solutions\pose.py", line 146, in init
File "mediapipe\python\solution_base.py", line 264, in init
FileNotFoundError: The path does not exist.
[15656] Failed to execute script 'mediapipepose' due to unhandled exception!
Press any key to continue . . .

Keep Failing to RE connect

It gets past the initial phase of connecting to PC Steam Vr but then it goes to this, every minute or two sending a new "Failed to connect"

Screenshot 2023-10-25 100012

Incorrect camera id

When I run it it says this:
Importing libraries....
Reading parameters...
Save file not found, will be created after you exit the program.
Opening camera...
Connecting to SteamVR
Camera opened!
Exception in thread Thread-2 (camera_thread_fun):
Traceback (most recent call last):
File "threading.py", line 1009, in _bootstrap_inner
File "threading.py", line 946, in run
File "mediapipepose.py", line 67, in camera_thread_fun
AssertionError: Camera capture failed! Check the cameraid parameter.
Starting pose detector...
INFO: Created TensorFlow Lite XNNPACK delegate for CPU.
Traceback (most recent call last):
File "mediapipepose.py", line 185, in
AttributeError: 'NoneType' object has no attribute 'copy'
[24024] Failed to execute script 'mediapipepose' due to unhandled exception!
Press any key to continue . . .

I have tried to change the camera id to practically everything in device manager and it does not work
Please Help!

Camera view showing just an orange screen

Everything goes well but when I press save and continue, the out window shows an orange screen
image

And when I set it to camera 2, it's a sky blue screen
image

if its because of the os, im using windows 11

Poor translation to steamvr

I'm not sure if I am the only one having this problem, but the 'out' window is tracking everything really well, but in SteamVR the movement is inaccurate and jittery, is there any way I can fix this?

[Request] Multiple Camera Setup

A Feature that allows you to add multiple cameras for better tracking at different angles would be very good. My room is quiet small and its hard to make due with just one camera as the FOV isn't high enough for me to move around. A feature to add multiple cameras and would help with this along-side tracking quality.
I'm not entirely sure how difficult this is to implement so sorry if this sounds like something that may be challenging to implement into this amazing software.

Much thanks for taking the time to read this,

  • Harvey

Error code -138

I tried to launch it and this occurred maybe I'm dumb but Idk

"Error in SteamVR connection"

Terminal says,
"Importing libraries....
Reading parameters...
Opening camera...
Connecting to SteamVR
Error in SteamVR connection. Retrying...
Camera opened!
Error in SteamVR connection. Retrying...
Error in SteamVR connection. Retrying...
Error in SteamVR connection. Retrying...
Error in SteamVR connection. Retrying...
Error in SteamVR connection. Retrying...
Error in SteamVR connection. Retrying...
Error in SteamVR connection. Retrying...
Error in SteamVR connection. Retrying...
Error in SteamVR connection. Retrying...
Could not connect to SteamVR after 10 retries!"

I have very little technical experience, sorry for the vague description.

Javascript implementation?

While the python implementation is ok right now. I can see this application getting more performance using the javascript API, i think it's easier to use the GPU as well

ip not loading in the web browser

Whenever i choose VRChatOSC, put in the oculus ip, enable webui, and click save and continue, i see the cmd and in it shows the ip i have to put in the browser but when i put it in the browser it just says error or could not load site. pls can someone help me.

Trying to use the OSC setting because SteamVR is not working for some reason, however I get this error message then it stops working entirely.

INFO: Start camera thread
INFO: Starting pose detector...
INFO: Created TensorFlow Lite XNNPACK delegate for CPU.
Traceback (most recent call last):
File "mediapipepose.py", line 181, in
File "mediapipepose.py", line 159, in main
File "backends.py", line 224, in updatepose
File "backends.py", line 156, in osc_build_bundle
File "backends.py", line 150, in osc_build_msg
File "pythonosc\osc_message_builder.py", line 159, in build
File "pythonosc\parsing\osc_types.py", line 231, in write_float
OverflowError: float too large to pack with f format
[32860] Failed to execute script 'mediapipepose' due to unhandled exception!
Press any key to continue . . .

This happens every time I try loading the application, anybody know why? Also a invite to the discord would be nice!

Not working with OBS Virtual Camera

Hello. I've been trying to use the application with OBS Virtual Camera since my Camera does not work standalone and have to stream the window the camera appears in. But whenever I start the application with OBS Virtual Camera this happens:
image

mediapipe puts a .dll file in my discord voice files/cameras randomly stopping working

first issue: was trying to get cameras back to working, and was going to delete all of the mediapipe files from my desktop to see if it was something with there being multiple versions, but it wasnt. windows popped up and said that a .dll file was open in my discord files, and now discord says that my files are corrupt.

second issue: as i stated above, my cameras just randomly stopped working. ive looked on other programs such as streamlabs, obs, and the actual camera app, and they all work there. ive reinstalled mediapipe and have restarted multiple times, to no avail.

Crash when enter VRChat from Oculus Quest 2, Virtual Desktop environment

Able to auto calibrate
Able to see the skeleton in SteamVR environment.
Stop executing when I click VRChat.

Python 3.9.2
Release 0.6

Log:
Postcalib y value: -83.56233936752928
Changed scale value to 1.0836985136298132
Traceback (most recent call last):
File "mediapipepose.py", line 243, in
IndexError: list index out of range
[26228] Failed to execute script 'mediapipepose' due to unhandled exception!

mediapipe vr and kinect 360. how work??

Hello everyone
I wanted to know if it is possible to use Kinect 360 with Media Pipe VR. I downloaded the latest beta version available, and when I select my kinect with the code "1" and start the program with Steam open, a totally black preview screen opens and I don't see the skeleton move inside. how can I make my kinect work with this tracking program? the only alternative I have left is amethyst but it doesn't work very well. Thanks for your help

Finger tracking

Using a phone strapped to a headset with a program to stream video to my computer as if it were a native camera, I can capture my hands.

Using this with mediapipe's finger and hand tracking program, It's feasible to add finger trackers to steamVR with just a phone.

How to use it on PC

I saw this video: https://www.youtube.com/watch?v=53JLgMyxi-4 the other day and tried it out but he doesn't explain what to do after you install it on PC, just for quest. Can someone please tell me how to use it on PC after I installed the drivers? Do I need a specific app or something? Thank you.

The program cannot find the phone.

The program cannot find the phone. I have set my computer and my phone on the same wifi, but no matter how I change the IP, the program cannot take effect.

Importing libraries....
Reading parameters...
Save file not found, will be created after you exit the program.
Opening camera...
Connecting to SteamVR
Error in SteamVR connection. Retrying...
Starting pose detector...
INFO: Created TensorFlow Lite XNNPACK delegate for CPU.
[tcp @ 0000018fff298f40] Connection to tcp://192.168.1.20:8080 failed: Error number -138 occurred
[ERROR:[email protected]] global D:\a\opencv-python\opencv-python\opencv\modules\videoio\src\cap.cpp (166) cv::VideoCapture::open VIDEOIO(CV_IMAGES): raised OpenCV exception:

OpenCV(4.5.5) D:\a\opencv-python\opencv-python\opencv\modules\videoio\src\cap_images.cpp:253: error: (-5:Bad argument) CAP_IMAGES: can't find starting number (in the name of file): http://192.168.1.20:8080/video in function 'cv::icvExtractPattern'

Camera opened!
Exception in thread Thread-2 (camera_thread_fun):
Traceback (most recent call last):
File "threading.py", line 1009, in _bootstrap_inner
File "threading.py", line 946, in run
File "mediapipepose.py", line 67, in camera_thread_fun
AssertionError: Camera capture failed! Check the cameraid parameter.
Traceback (most recent call last):
File "mediapipepose.py", line 185, in
AttributeError: 'NoneType' object has no attribute 'copy'
[21700] Failed to execute script 'mediapipepose' due to unhandled exception!

Hand tracking on android over wifi?

One thing I'd love to see is the ability to track your hands on VR with a phone. You could theoretically track your hands in real-time by strapping your phone onto your headset, run mediapipe hand-tracking on it, and broadcast the positional data over Wi-Fi, then use your tracker to receive the data and inject it into steamVR

camera shows up black on preview and doesnt track

it worked the first time i ran it with my usb camera on 0. when restarted it, it was only able to start on 1 and the screen showed up black

i have
deleted the software and re downloaded it

checked the camera to make sure it was working

checked camera privacy settings

deleted camera from device and plugged it back in

tried to use v0.6

unplugged the camera and plugged it back in

and restarted my pc multiple times

im not sure what to do to fix it

I have a trouble in installation activateMultipleDrivers

Steam directory found. (C:/Program Files (x86)/Steam)
SteamVR directory located. (C:/Program Files (x86)/Steam/steamapps/common/SteamVR)
Traceback (most recent call last):
File "installer.py", line 60, in
File "json_init_.py", line 293, in load
UnicodeDecodeError: 'gbk' codec can't decode byte 0xac in position 959: illegal multibyte sequence
[11200] Failed to execute script installer

[已退出进程,代码为 4294967295 (0xffffffff)]

multicamera setup

is there any way to take data from more than one camera? I've run into issues with one limb obscuring another or simply standing sideways in front of the camera where I lose tracking, and I'd like to know if there's any way to connect a second webcam to Mediapipe to fill in for these situations.

-- Subject

Support multi camera ?

So there is a similar software called freemocap. It's also open-source the problem is that it is not quite real-time and doesn't support vr. But it supports multi-camera and just needs to calibrate using some sort of apriltag marker at once. I've wondered if this can also support multi-camera. I know it will take a lot of computer resources. Maybe add GPU acceleration? (I know in the current stage it is better to implement somethings that can increase user experience but maybe this can be done in more further the future?)

Runs at 70fps, but is very laggy in VR.

Mediapipe Pose runs SUPER smooth with my setup with little to no jitter and perfect accuracy. However, this is not the case when in VR. My movements are delayed by 100-150ms (ok I can deal with that no problem) but the trackers seem to be moving around somewhere between 20-30fps and jitter a LOT. I have played with the smoothness settings and while they do help, I can't seem to find an inbetween for smooth tracking and high accuracy. What confuses me the most is that Mediapipe Pose hardly even seems to be jittering at all in the first place!

Error Connecting To Steam VR

I have checked the driver and it is installed and it shows up in the addons page. However, When I check The Logs I see :

Sat Jan 08 2022 15:43:17.825 - Unable to load driver apriltagtrackers from C:\Program Files (x86)\Steam\steamapps\common\SteamVR\drivers\apriltagtrackers\bin\win64\driver_apriltagtrackers.dll (126)

Sat Jan 08 2022 15:43:17.825 - ASSERT: "DriverManager apriltagtrackers load error 126u" at c:\buildslave\steamvr_rel_win64\build\src\common\vrcommon\drivermanager.cpp:405.

Sat Jan 08 2022 15:43:18.768 - Unable to load driver apriltagtrackers because of error VRInitError_Init_FileNotFound(103). Skipping.

I have no idea what is going on.
Please Help

Trackers Not Working

so i installed the drivers, i ran the start_mediapipepose.bat i configured the settings, i pressed Save And Continue, 2 windows appear and after 2 seconds they dissapear. SteamVR Appears To Have (I Think) 6 Tracker Icon's, which all say "Move To Enable Tracker" (smthing like that). When I Go Try A Game That Supports Tracker's, They Don't Work. Ill Now Show Some Settings
image

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.