stereo images with ground truth disparity for machine learning

Описание: The 3d Camera collects stereo pair data with ground truth disparity images.The project came about because I couldn't find any decent data to use in my project, so I decided to generate some myself. To demonstrate, I train a convolutional neural network to extract the disparity image from the stereo pair. The camera is made from a Synchronized Dual Lens Stereo USB Camera and an Arducam Time of Flight camera. They are both connected to a raspberry pi.

Equipment:
Raspberry pi 4
Arducam Time of Flight Camera: see for example: https://core-electronics.com.au/time-...
Stereo camera: Camera USB-960P2CAM I bought mine from Amazon, but best to shop around. link is https://www.amazon.com.au/gp/product/...

Links:
Previous video on arducam time of flight camera:    • Arducam time of flight camera for 3d segme...  
Arducam getting started https://docs.arducam.com/Raspberry-Pi...
Arducam github: https://github.com/ArduCAM/Arducam_to...
Building a CNN based autoencoder:   / building-a-cnn-based-autoencoder-with-deno...  

Chapters:
0:00 Introduction
1:26 Camera and equipment
2:17 Demonstration
6:30 Portable camera for data recording
7:06 Stereo image pairs and disparity data recordings
8:41 Code used on raspberry pi
10:32 CNN outline
14:21 CNN pyhone notebook explanation.