Test Real-time 3D Reconstruction

[chenchen2015]

Description
Use the trained neural networks model from #13 to test reconstruct new gantry positions and evaluate static and dynamic reconstruction accuracy.

TODO

• Write a MATLAB script that constantly captures camera views and processes the images to find the target (outputs Nx2 matrix that contains 2D image coordinates of N camera views)
• Test the MATLAB script. Find a gantry position range that we can have a reliable output.
• Feed the output of the image processing function to the trained neural network function to get the reconstructed gantry coordinates.
• Read Gantry ground truth and compare results
• Use the result to estimate static reconstruction accuracy - reconstruct N times at a given location and use the sample distribution to estimate the mean and confidence interval
• Estimate dynamic reconstruction accuracy by moving the gantry unit distance away