3D Scanning algorithm with librealsense
Budget: €3,000 – €5,000 EUR
Brief description:
An existing C++ scanning algorithm is leveraging librealsense SDK to acquire the 3D face of a person in front of it.
The person moves the head to the left and right, and the algorithm acquires the point-cloud and images, stitches them and outputs a .ply 3D model.
The goal of the project is to improve the existing algo, specifically
- Introducing face recognition based on landmarks:
--->to improve the overall quality of the output (currently the algo is basing its calculations on a set distance from of the subject from the camera and estimates a rotation for the neck/head in the different directions)
--->to improve the visual feedback by showing which areas of the face have been scanned and which ones are to be scanned yet
- Introducing better feedback for the operator during the scan (i.e.: subject is moving too fast/is too close, etc).
- Introducing GPU computation to speed up the calculation, instead on full CPU
- Introducing the possibility to scan also the neck and hair when the person moves the head up and down during the scan
- Introducing the possibility of scanning the person when he/she has something in the mouth (a calibration object with some reference shapes)
- More details will be handed over after acceptance.
We are open to modification of the existing algo, or a brand new algorithm (C++, Python, etc) that should run on Windows as a dll/module of a WPF/C# existing software.
An existing C++ scanning algorithm is leveraging librealsense SDK to acquire the 3D face of a person in front of it.
The person moves the head to the left and right, and the algorithm acquires the point-cloud and images, stitches them and outputs a .ply 3D model.
The goal of the project is to improve the existing algo, specifically
- Introducing face recognition based on landmarks:
--->to improve the overall quality of the output (currently the algo is basing its calculations on a set distance from of the subject from the camera and estimates a rotation for the neck/head in the different directions)
--->to improve the visual feedback by showing which areas of the face have been scanned and which ones are to be scanned yet
- Introducing better feedback for the operator during the scan (i.e.: subject is moving too fast/is too close, etc).
- Introducing GPU computation to speed up the calculation, instead on full CPU
- Introducing the possibility to scan also the neck and hair when the person moves the head up and down during the scan
- Introducing the possibility of scanning the person when he/she has something in the mouth (a calibration object with some reference shapes)
- More details will be handed over after acceptance.
We are open to modification of the existing algo, or a brand new algorithm (C++, Python, etc) that should run on Windows as a dll/module of a WPF/C# existing software.