Closed dippmatt closed 2 months ago
The BOP challenge renderings are used for training. This conversion is done. Issue can be closed as soo as the validation data, the Canon Images, are converted too.
Several things to consider: The BOP dataset is enough for training, validation and testing. Use the "BOP'19-23 test images" for test & val. These are from the Primesense camera. Use "Synt. training images of isolated object" for training
Convert to yolo all format
Download and convert annotations of T-Less dataset to work with YoloV8.
Toolkit to render 3D objects from scenes: https://github.com/thodan/bop_toolkit