Hi, thanks for sharing the dataset!
I noticed that contact.json contains 39,154 tactile frames, but when I load the training set using the code, the actual number of training samples seems smaller(38605).
Could you please clarify how many in-contact samples are actually used for training? Is there any filtering or sampling involved?
Hi, thanks for sharing the dataset!
I noticed that
contact.jsoncontains 39,154 tactile frames, but when I load the training set using the code, the actual number of training samples seems smaller(38605).Could you please clarify how many in-contact samples are actually used for training? Is there any filtering or sampling involved?