xji-apex

same ...

嗯嗯可以了，谢谢~

同样的问题，有人可以指点一下嘛？

 

same.....

Thank you for your timely reply and update! Now the problem has been solved, but I find that the actual pixel displacement is twice or more than the predicted result....

Yeah, - The dataset is the ([shapes_translation](https://rpg.ifi.uzh.ch/davis_data.html) ). I use the optical flow value of a corner to predict its pixel position and compare it with the actual pixel position...

I use the image_watch tool to check the corner position and the predicted optical flow value. As you said, events have a great correlation with the camera moving. I'll test...

same error