Open z-yf17 opened 1 year ago
Thank you for your interest in our work. The position of the object can be inferred from the 2d coordinate map (x2d) of that cropped image, which include coordinates in the original image space. These coordinates are used as the input of the PnP layer. Besides, in EPro-PnP-6DoF v2,we also improved the scale handling by feeding the original size of the cropped image. Hope these can answer your question.
Hi, thanks for your great work!
I noticed that your input image is cropped before training and inferencing. After obtaining the inferencing results, I can accurately obtain the rotation angle of the object, but the position of the object is relative to the cropped image. How can I obtain the relative position between different objects? Or, how can I know the absolute position of each object in the entire image?
Looking foward to your reply!