Hello,
Thank you for your excellent and impressive work. We believe that your innovative work will be excellent in many downstream tasks.
Thus, I wanted to take the chance and ask you if, and if yes, how should we make customized datasets based on our application scenarios?
Take S3DIS dataset format as an example,
We noticed that your MIT model does not rely on camera pose and depth map, so are there any specific requirements for the point cloud data and images to be collected during the data acquisition phase? For example, does the image data have to be acquired at the same moment while the point cloud data is acquired? Is there any requirement on the amount of image data?
Since S3DIS datdset is already done with point-wise and pixel-wise labeling, but how do we make scene-level labels in a custom task?
If all goes well and I have trained MIT on a custom dataset based on my application scenario, can I still generate segmentation labels during the predictive inference phase if I only have point cloud data?
Please let us know if there are other considerations for making custom datasets.
I know your time is very valuable, but we would still really like to have your answers, thank you in advance!
Hello, Thank you for your excellent and impressive work. We believe that your innovative work will be excellent in many downstream tasks. Thus, I wanted to take the chance and ask you if, and if yes, how should we make customized datasets based on our application scenarios? Take S3DIS dataset format as an example,
I know your time is very valuable, but we would still really like to have your answers, thank you in advance!
Best Wish！