Closed KeyPisces closed 4 years ago
Hi @KeyPisces, you can see in data_loader.py that the main items the dataloader returns are: RGB image, back-projected depth image (XYZ), foreground labels, and direction labels. The dimensions of each tensor are documented in the comments. Please take a look at the code for more details.
Unfortunately I cannot provide my training dataset (TOD) at this time. However, OID is a public dataset and you can download the segmentations. You will probably have to do a bit of processing to get them in the form that is useable from the dataloader, or you can just edit the dataloader.
Hello, first of all thanks for sharing your amazing job. I want to know What are the main types of things on your TOD dataset! and if you can provide the dataset you used for training (TOD and set of OID data) and testing. Thanks in advance ;)