Open feivellau opened 2 days ago
+1 Observed the same. Also the full number of files should be 116k given what is stated in the paper
please check this reply for details. https://github.com/bytedance/coconut_cvpr2024/issues/32#issuecomment-2458505592
Hi, I tried creating the COCONut-Large dataset and noticed the number of annotations were different than expected. I got panoptic_object365.tar and panoptic_objects365train.json on huggingface. But, the amount of data is not right.