Can anyone help me regarding this problem Search n°1 : we need to find a tool capable to convert an image (a real photo of a naked object on white background) into another image (virtual photo) of the same object but being seen from a different angle. The input to this tool is one photo of an object on white background being shot from a point a view that is what it is. The output of this tool is one photo of the same object but being seen for a wanted different angle and position of the camera in space.
What is the difficulty ? The difficulty is because some areas of the object will be hidden in the first single provided photo, so the tool must be clever enough to “imagine” what could be the shapes and the look like of these hidden parts.
At the end what we need is a tool with :
- Inputs to the tool: o the original single image that we have of the object on a white background, o the position of the camera versus the tool (it is our specification) that we wish.
- Outputs of the tool: o the new photo (hat will be virtual) of the object according to the specified desired angle but still on white background. o eventually the 3D model of the object that will have been done by the tool that will have “imagined” what are the shapes of the hidden parts of the object in the first photo. I have written “eventually”
We have tired all the AI tools for this finally I found an article like deep implicit surface network I like it quit okay I attach the paper link: https://github.com/laughtervv/DISN
But I have a doubt on creating the dataset which tool I have to use for this type of annotation and someone can help with relvalent paper means it will usefull for me