What is the accuracy of the 3D contour image reconstructed from two or abover 2D images?
caibf opened this issue · comments
According to your publicly available paper, the average imaging accuracy is 2.7mm in the absence of specific camera parameters. What is the accuracy we can achieve if we use a high-resolution 2D camera to perform 3D reconstruction within a straight distance of about 15cm?
I guess it should be similar to the performance of the DTU dataset, since the setup would be roughly equivalent. This really depends on the observation setup, so you would need to quantify the quality in your particular setup. Keep in mind that without known GT cameras, the reconstruction is done up to an unknown scale factor.
I guess it should be similar to the performance of the DTU dataset, since the setup would be roughly equivalent. This really depends on the observation setup, so you would need to quantify the quality in your particular setup. Keep in mind that without known GT cameras, the reconstruction is done up to an unknown scale factor.
Thank you for your valuable suggestion.
I haven't practiced this open source library yet. Would it be convenient for you to recommend relevant hardware configurations or models for me to refer to? For example, 2D camera configuration or related parameters.
You can try the model with a simple phone camera and see it's strengths and weaknesses. In our experience you will get the best quantitative results with minimal distortion lenses (model was trained on undistorted images). As for the camera setup, this is really application dependent so I can't tell you beforehand. Luckily it is fast to iterate over setups if you have only a few images.