InvPT: Inverted Pyramid Multi-task Transformer for Dense Scene Understanding

Open paper

Benchmark	Model	Rank	Results
boundary-detection-on-nyu-depth-v2	InvPT	#1	odsF: 78.1
semantic-segmentation-on-nyu-depth-v2	InvPT	#31	Mean IoU: 53.56%
monocular-depth-estimation-on-nyu-depth-v2	InvPT	#66	RMSE: 0.5183