Are you running into an issue or bug not described below? Let us know by reporting a bug and we will address the issue as soon as we can!
How do I reduce the number of dropped frames?
For a complete guide on increasing capture performance, see the Verify Computer Performance page.
Color Data Shadow
In the 3D view-port, you may notice a slim outline in your Depthkit capture when the depth rangedepth range - The depth near and far planes represented as hue encoded ranges, red as the near plane and pink as the farthest plane before clipping. is set to the maximum range. This is a characteristic of the depth sensor that can appear like a double image or outline. This behavior can be described in the image below.
Imagine the two dots as spheres on the same plane. The two lines represent single pixels, or rays, in the camera (blue in depth, and red in color). Notice that on the point where the blue sphere is detected in the depth camera, the path to the color camera actually passes through the red sphere in the foreground. This means when the color on that point of the background sphere is displayed, it will actually receive the color from the foreground sphere. If you think of the color camera as a "data light", then the background sphere is in the "color data shadow". The foreground sphere is eclipsing it's color data. When we look from the perspective of the depth camera, we see a double image of the occluded data.
There are holes in my depth data.
Due to the way in which depth cameras sense depth, there are certain materials that cause errors in the depth data during capture. These are usually reflective or shiny materials that will cause holes and similar artifacts in your depth stream. Read more about material and garment constraints. To repair and reduce these artifacts, Depthkit's Refinement algorithm is designed to clean up and enhance your volumetric data.
My takes are no longer visible in my Depthkit project.
- This is likely due to a file that may be missing from your project folder. Ensure that if you relocate your project to a new computer or drive, that you move all files and folders within the Depthkit project parent folders.
Do not rename files within your Depthkit project. The dkproject.json is the source of truth of all of your project information. If your take folders are renamed, the dkproject.json file will no longer be able to read the takes when you launch the project in Depthkit. Instead, name the take from within Depthkit before your capture each take.
Do not move take folders outside of your Depthkit project folder. If you relocate takes outside of the Depthkit parent folder, the new file path will not update automatically and Depthkit will not be able to read the takes when you launch the project.
Do not delete the _calibration folder or dkproject.json file. Both files hold essential information about your project and should remain untouched within your Depthkit project folder.
When opening an existing project, my project does not open and takes a long time to load.
Make sure you are selecting the parent project folder. Do not select and individual take within that folder.
If you have relocated your project folder, make sure it is not opened from a exFat drive. If you need to move your project to an external hard drive, use the NTFS format.
When opening an existing project, I get an error message saying "Application has requested more GPU memory than is available"
When opening an existing project with many recorded takes, you may get an error that closes Depthkit. To work around this:
Make a copy of your project folder (complete with all of the contents).
Remove some of the take folders from the new copy.
Open Depthkit and open this pared-down project folder instead of the original.
The clip names of the removed takes will still appear in the Edit context's Library panel, but show red alert icons indicating that the clips are offline. To restore those clips, simply add the take folder to the new project folder and restart Depthkit.
Console Error Messages
Color frame dropped. Depth frame dropped.
Dropped frames will be reported in the Depthkit console, however, you should only watch for dropped frames in the diagnostics panel above the record button that becomes visible when recording. If this panel is reporting dropped frames, follow the steps to verify computer performance. This diagnostics panel will be your source of truth, so you don't need to worry about watching for performance issues in the console window.
Console Warning Messages
Failed to load legacy take: unable to determine asset version.
The error above may be present if you are importing a legacy project into Depthkit. It means that you likely have folders within your Depthkit project that are not take folders. This is fine and will not have a negative impact on your project. You can disregard this message.
Azure Kinect Issues
Depth & Color Alignment Inaccuracy
On some Azure Kinects, we have noticed a slight misalignment between the sensor depth and color, especially when more than 3 feet from the sensor. We are currently investigating this issue. For more details, check out the Azure Kinect SDK github.
"Failed to open device" when I plug in my Azure Kinect.
There could be a connectivity or power issue with your sensor.
- Verify that your depth sensor is plugged in and well powered. You can check this via the power indicator LED on the back of the device.
- Restart your computer and relaunch Depthkit once you have checked that your sensor is properly powered.
Kinects for Windows v2 Issues
My Kinect is not being detected by Depthkit.
1) Ensure that your sensor is securely plugged in and powered on.
If it is a power issue, you may want to try plugging it directly into the power outlet in case an extension cable is causing insufficient power.
2) Ensure that you have installed the Kinect SDK.
If you have not done so, Depthkit will launch, but you will not be able to see depth or color streams.
3) Check the performance of your sensor with the Kinect Configuration Verifier Tool.
This will give you insight on the state of your sensor connection. Ideally, all categories will be checked and green.
Make sure you have your sensor connected to your computer via a USB-3.1 / USB-C (gen 2) port. Gen 2 ports have a much higher data transfer rate and will promote a more stable connection and data stream from your depth sensor. If you computer does not have a USB 3.1 (gen 2) port, try a USB to USB-C adapter. Learn more about USB ports and how to find which one you are currently using.
You cannot run the Kinect Configuration Verifier while you are running Depthkit, as you will find that the depth/color streams may not show up in Depthkit when both apps are active. With this in mind, once your Kinect performance is confirmed, close the verifier before you open Depthkit.
My Kinect is repeatedly disconnecting from the computer.
First, update your WDF Kinect Sensor Interface driver. You can find this in the Device Manager, under Kinect sensor devices. In the drop down, right-click on WDF KinectSensor Interface and select update driver.
If this does not solve the issue, explore the Verify Computer Performance page.
Depthkit is displaying a low performance notice in the Record Panel.
To solve for under-performance, follow the instructions in Verify Computer Performance.
My Kinect is capturing at 15 fps instead of 30 fps.
If there is not enough light in your scene, the Kinect fps may drop to 15 fps. To avoid this, make sure your capture area is well lit. You will know when this happens if the Depthkit console reports the warning "low light activated". This low light mode will be deactivated when lighting is increased.
Depthkit is not detecting my subject during capture when capturing with the Kinect for Windows v2.
Depthkit detects a body or bodies based on body indexbody index - The detection of one or multiple bodies in the depth data. When generating your assets, this acts as a body mask, allowing you to simply hide the background for clean exports of your subjects. This eliminates the need to shoot on a green screen. data.
Make sure the full body is in the frame. The body indexbody index - The detection of one or multiple bodies in the depth data. When generating your assets, this acts as a body mask, allowing you to simply hide the background for clean exports of your subjects. This eliminates the need to shoot on a green screen. cannot detect a partial body or a subject that is being obstructed by an object.
Make sure the subject is not too close to the sensor. The Kinect will not detect a body if too close to the sensor.
- Ensure that your Kinect is level and not rotated. You may be tempted to capture vertically as you would with a DSLR. However, the sensor cannot detect bodies if rotated or in a vertical position.
RealSense Depth Camera Issues
My color stream appears to be flipped or rotated compared to the depth stream.
To solve this issue, update the firmware on your RealSense to the latest version. Find information on firmware releases and how to update here.
There are artifacts in my combined per pixel clip in Unity, located at the near plane of my subject, that are not visible in the Depthkit application.
This currently a known issue related to the far plane value of your depth range, when the far plane is very close to your subject, as seen in the image below. It applies to combined per pixel exports in Unity when using the Unity video player as your Depthkit clip video player.
To avoid this Unity artifact,
Move the far plane away from your subject slightly, so the hue encoded depth is no longer red.
If you have already exported your clips and do not have time to re-export, set your Depthkit clip video play to AVPro in Unity. This will get rid of any unwanted artifacts.
Updated about a month ago