Is the test results normal?

[pyni]

Hi~ I use your training weights (ape_pvn3d_best.pth.tar) to test some results. For the real images, the test results are quite awesome:

But for rendered images, the results are shown as follows: For rendered images with real background, the results are shown as follows:

I am not sure whether the results are normal? And does your training weights( pvn3d_best.pth.tar) use rendered images for training?

[ethnhe]

• The training weights use rendered images for training.
• We didn't inference on the synthesis data. But since the validation set only contains real scene images, It's possible that the selected weights perform better on real data than synthesis data.