What’s the best way to decode what I think is referred to as either compressed or compact RLE format (for example COCO-Stuff segmentation annotations)? It doesn’t look like rle_decode supports this.
Example segmentation annotation from COCO-Stuff:
{'counts': 'eh]52U=5L5K3L5YC]Oa<h0M4K4M3L4M4M2000001O01O0000000000000001O000O15K8H8H7I8H8FbR9d0llFj0VOj0VOi0WO8H0000000001O0001O0000000000J6C=ZOf0\\OkDdNc;m0e0TOeC6m<Jc]k1', 'size': [427, 640]}