4 ms·

Anyone has an interesting project to build with Kinect? I have a 360 kinect laying around, since nobody does Just Dance parties anymore. Also, was the XBone ver

by merpkz 2y ago

Anyone has an interesting project to build with Kinect? I have a 360 kinect laying around, since nobody does Just Dance parties anymore. Also, was the XBone version of kinect much better than 360?

knd775 2y ago
The 360 one is better for 3d scanning. Xbox one version is better for object tracking.
- forgotacc240419 2y ago
  Isn't this just because there was nowhere near as much software for 3d scanning made for v2?
  - numpad0 2y ago
    There was something unsatisfactory to V1 users in the V2 system. Depth was jittery or whatever it had been. V1 was based on an Israeli structured light tech that later became Apple TrueDepth, a static dot pattern taken with a camera which deviations becomes depth. V2 was technically an early and alternative implementation of Time-of-Flight LIDAR based on a light phase delaying device, that output deviations from set known distance as pictures or something along that. There wouldn't have been lacking app support issues if V2 did work. There was/were something that made users be "yeah... by the way" about it.
    - sleepybrett 2y ago
      The problem with the v2, from someone who had a boss that loved the kinect and used it in retail environments experimentally wasn't the new tech. The new tech was/is amazing. The level of detail you can get from the v2 dwarfs the v1 on every axis. The problem was that it ONLY had a windows SDK and most of the people who did amazing work outside of games and the xbox with the kinect v1 were using it with macs and linux in tools like openframeworks and processing. The v1 was developed outside microsoft and primesense and there were SDKs that were compatible cross platform. Tons of artists dove right in. The KinectV2 only offered a windows sdk, and that's what killed it in the 'secondary market'. Luckily now we have libfreenect2 (https://github.com/OpenKinect/libfreenect2 https://github.com/OpenKinect/libfreenect2) but that came too late for most of us.
      - forgotacc240419 2y ago
        Interesting stuff, thanks for the clarification! Did MS provide SDKs for other OSes for v1 or was it just easier for people to make an open source one?
        sleepybrett 2y ago
        primesense made them, the guys who sold the sensor to microsoft.
piva00 2y ago
If you would like to venture into generative/interactive visuals there are libraries for TouchDesigner which uses the Kinect. An example: https://www.youtube.com/watch?v=GDZoOnzLYGo https://www.youtube.com/watch?v=GDZoOnzLYGo
xandrius 2y ago
In academia it's often used as a cheap environment scanner, accessibility studies for people with limited mobility, etc. I worked on a few projects where the Kinect was mounted on a self-driving robot for 3D mapping of areas unaccessible to humans. Also worked in using it in applications with people with Alzheimer's. The issue is very often lack of or outdated SDKs for the target platform. The hardware could be better but it's already getting many project to 90%.
rednab 2y ago
the Kinect One is better in a bunch of ways (field of view, resolution) but a big one for certain use-cases is that it can fully track 6 skeletons. The 360 Kinect can only track two skeletons (but differentiate 6).
- CYR1X 2y ago
  Source on that tracking 6 skeletons? That's cool.
lelag 2y ago
The problem with the original Kinect (v1) is that good tracking software for it was never really written. Most application that support it, just use the original Microsoft SDK to do the motion tracking, and it's just not very good, the main issue is that it always assume that the tracked person is directly facing the camera, and is very bad at dealing with occlusion. The good thing about it, is that it ran in real-time on a potato. In order to get a good result, someone would need to train a good model for HPE that could use the cloud point data directly, but it seems nobody cares about depth sensor anymore, most efforts are going to HPE from regular 2d video (like media pipe holistic model). And given the result you can get with media pipe, openpose and the likes, it's understandable nobody is bothering working with low resolution cloud point anymore for 3D HPE. The only use-case I can think of for a Kinect v1 in 2025, would be robotics if you want a low latency low resolution cloud point for your robot control, but even there I think we are moving to big vision model capable of making sense of regular video feeds.
- Fethbita 2y ago
  There might be some work at https://k2vr.tech https://k2vr.tech regarding this it seems like.

by merpkz 2y ago

Anyone has an interesting project to build with Kinect? I have a 360 kinect laying around, since nobody does Just Dance parties anymore. Also, was the XBone version of kinect much better than 360?

knd775 2y ago

The 360 one is better for 3d scanning. Xbox one version is better for object tracking.

forgotacc240419 2y ago
Isn't this just because there was nowhere near as much software for 3d scanning made for v2?
- numpad0 2y ago
  There was something unsatisfactory to V1 users in the V2 system. Depth was jittery or whatever it had been. V1 was based on an Israeli structured light tech that later became Apple TrueDepth, a static dot pattern taken with a camera which deviations becomes depth. V2 was technically an early and alternative implementation of Time-of-Flight LIDAR based on a light phase delaying device, that output deviations from set known distance as pictures or something along that. There wouldn't have been lacking app support issues if V2 did work. There was/were something that made users be "yeah... by the way" about it.
  - sleepybrett 2y ago
    The problem with the v2, from someone who had a boss that loved the kinect and used it in retail environments experimentally wasn't the new tech. The new tech was/is amazing. The level of detail you can get from the v2 dwarfs the v1 on every axis. The problem was that it ONLY had a windows SDK and most of the people who did amazing work outside of games and the xbox with the kinect v1 were using it with macs and linux in tools like openframeworks and processing. The v1 was developed outside microsoft and primesense and there were SDKs that were compatible cross platform. Tons of artists dove right in. The KinectV2 only offered a windows sdk, and that's what killed it in the 'secondary market'. Luckily now we have libfreenect2 (https://github.com/OpenKinect/libfreenect2 https://github.com/OpenKinect/libfreenect2) but that came too late for most of us.
    - forgotacc240419 2y ago
      Interesting stuff, thanks for the clarification! Did MS provide SDKs for other OSes for v1 or was it just easier for people to make an open source one?
      - sleepybrett 2y ago
        primesense made them, the guys who sold the sensor to microsoft.

piva00 2y ago

If you would like to venture into generative/interactive visuals there are libraries for TouchDesigner which uses the Kinect. An example: https://www.youtube.com/watch?v=GDZoOnzLYGo https://www.youtube.com/watch?v=GDZoOnzLYGo

xandrius 2y ago

In academia it's often used as a cheap environment scanner, accessibility studies for people with limited mobility, etc. I worked on a few projects where the Kinect was mounted on a self-driving robot for 3D mapping of areas unaccessible to humans. Also worked in using it in applications with people with Alzheimer's. The issue is very often lack of or outdated SDKs for the target platform. The hardware could be better but it's already getting many project to 90%.

rednab 2y ago

the Kinect One is better in a bunch of ways (field of view, resolution) but a big one for certain use-cases is that it can fully track 6 skeletons. The 360 Kinect can only track two skeletons (but differentiate 6).

CYR1X 2y ago
Source on that tracking 6 skeletons? That's cool.

lelag 2y ago

The problem with the original Kinect (v1) is that good tracking software for it was never really written. Most application that support it, just use the original Microsoft SDK to do the motion tracking, and it's just not very good, the main issue is that it always assume that the tracked person is directly facing the camera, and is very bad at dealing with occlusion. The good thing about it, is that it ran in real-time on a potato. In order to get a good result, someone would need to train a good model for HPE that could use the cloud point data directly, but it seems nobody cares about depth sensor anymore, most efforts are going to HPE from regular 2d video (like media pipe holistic model). And given the result you can get with media pipe, openpose and the likes, it's understandable nobody is bothering working with low resolution cloud point anymore for 3D HPE. The only use-case I can think of for a Kinect v1 in 2025, would be robotics if you want a low latency low resolution cloud point for your robot control, but even there I think we are moving to big vision model capable of making sense of regular video feeds.

Fethbita 2y ago
There might be some work at https://k2vr.tech https://k2vr.tech regarding this it seems like.