Open NTUZZH opened 1 year ago
Hello,
We are using a RaycastLaser https://github.com/kolaszko/hl2_detection/blob/alpha/MachineLearningPlayground/Assets/Scripts/RaycastLaser.cs from https://github.com/EnoxSoftware/HoloLensCameraStream to find a collision of a ray with a spatial map and use the collision point as a position of a 2D bounding box.
Cheers, Mikołaj
Thanks for your reply!
Hi there,
It's my honor to read your latest paper about deploying the YOLO_v8 object detection model with HoloLens2.
I have one query about back-projecting the bounding box from 2d to 3d. Assuming that we already get the output from YOLO model, which is a 2d image with detected objects and bounding box. Then we want to project this bounding box into the precise location of the 3D world in the HoloLens2's view. To realize this, we need to find the 3D coordinate of this bounding box to place it. My question is how can we get a 3D coordinate from just a 2D image from the model output? Shall we use the depth information provided by HoloLens2?
I will be so appreciate if kindly you could discuss this problem with me! ♥
Regards, -Ziheng