Closed zhang-qihao closed 1 month ago
Hi @zhang-qihao,
Thank you for your interest in our work! We are planning to release the dataset with no image augmentation, which should reduce the dataset size to 1/7 in theory. We are also releasing the raw RGB images of the dataset soon, so you can use them to build a smaller version of the dataset.
Hope this helps!
Hi @zhang-qihao,
We uploaded raw RGB images (with corresponding depths and object masks), so you can use them to build a smaller version of the dataset (e.g., using no augmentation, using only egocentric images, etc.). Let me know if you have any further questions.
The 1.6T dataset is too big, is there a smaller one that allows me to train the model quickly?