Hello. First of all, thank you for your great project.
I was recently impressed by your NVS model, so I tried your NVS with my own data.
But when I saw the results, a question arose. There is clearly only one person in the single image I entered as input data, but two people were displayed in all result frames.
I understood that your NVS model first creates a 3D object from a single image and then divides it according to the camera direction to create the NVS.
Because it is a single image, it is understandable that points are created beyond the back boundary of the object. However, looking at the resulting frame, it appears that two people were created independently of each other.
If this result is caused by a problem with object boundaries (especially the back boundary of the object) caused by a single image, shouldn't the two people be connected to each other?
I wonder if it's because of filtering. Is there a way to check the 3D objects created during the sv3d_p process?
Hello. First of all, thank you for your great project.
I was recently impressed by your NVS model, so I tried your NVS with my own data. But when I saw the results, a question arose. There is clearly only one person in the single image I entered as input data, but two people were displayed in all result frames.
I understood that your NVS model first creates a 3D object from a single image and then divides it according to the camera direction to create the NVS.
Because it is a single image, it is understandable that points are created beyond the back boundary of the object. However, looking at the resulting frame, it appears that two people were created independently of each other. If this result is caused by a problem with object boundaries (especially the back boundary of the object) caused by a single image, shouldn't the two people be connected to each other?
I wonder if it's because of filtering. Is there a way to check the 3D objects created during the sv3d_p process?
Thanks!