From b74d1fdc6b57eb03f1b03286d6ff4369a3ba562f Mon Sep 17 00:00:00 2001 From: PradeepKadubandi Date: Mon, 27 Mar 2023 07:18:16 -0700 Subject: [PATCH] This was probably a simple oversight? I think the entry for ADE20K in the comments was added by mistake in the middle of description for WOD-PVPS making it confusing for the reader of documentation. --- data/dataset.py | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-) diff --git a/data/dataset.py b/data/dataset.py index 7f8c4fe..aab725e 100644 --- a/data/dataset.py +++ b/data/dataset.py @@ -58,13 +58,6 @@ contains high quality panoramic video annotations with time and cross-camera consistency. -8. ADE20K panoptic segmentation -(https://groups.csail.mit.edu/vision/datasets/ADE20K/) - -Panoptic segmentation annotations for ADE20K dataset. Note that we convert the -provided ADE20K panoptic segmentation format to the following one: -panoptic label = semantic label * 1000 + instance id. - We can use the dataset in the following settings: - In the multicamera setting, an example contains all camera data within a frame (i.e., the instance correspondence between cameras are guaranteed). @@ -86,6 +79,13 @@ users could ignore the provided depth groundtruth and use the dataset for the task of panoramic video panoptic segmentation. +8. ADE20K panoptic segmentation +(https://groups.csail.mit.edu/vision/datasets/ADE20K/) + +Panoptic segmentation annotations for ADE20K dataset. Note that we convert the +provided ADE20K panoptic segmentation format to the following one: +panoptic label = semantic label * 1000 + instance id. + References: