wenbowen123 / iros20-6d-pose-tracking

[IROS 2020] se(3)-TrackNet: Data-driven 6D Pose Tracking by Calibrating Image Residuals in Synthetic Domains
Other
384 stars 66 forks source link

produce_train_pair_data.py clarificaiton #80

Open DanieleBertagnoli opened 6 months ago

DanieleBertagnoli commented 6 months ago

Hi, I was trying to understand what the produce_train_pair_data.py script does. Using the blander_main.py we have generated the synthetic images of the objects, then we have to call produce_train_pair_data.py to generate pair data. What do you mean with pair data?

wenbowen123 commented 6 months ago

During training, the networks takes input a pair of image to learn the relative pose between them. I'd also recommend to watch our recent work FoundationPose which does not need any training and can be applied immediately.