notesum.ai

Published at December 5

MT3DNet: Multi-Task learning Network for 3D Surgical Scene Reconstruction

cs.CV

cs.AI

cs.HC

cs.LG

Released Date: December 5, 2024

Authors: Mithun Parab¹, Pranay Lendave², Jiyoung Kim³, Thi Quynh Dan Nguyen⁴, Palash Ingle³

Aff.: ¹R.J. College, Mumbai, India; ²K J Somaiya College of Engineering, Mumbai, India; ³Sejong University, Seoul, South Korea; ⁴Korean Institute of Oriental Medicine, Daejeon, South Korea

Arxiv: http://arxiv.org/pdf/2412.03928v1

Model	Regime	Arch.	Segmentation	Object detection
			Dice	mAP
TernausNet11[8]	Seg.		0.920
ERFNet[19]	Seg.	ERFNet	0.923	-
LinkNet[3]	Seg.	UNet	0.941	-
DeepLabv3+[2]	Seg.	ResNet	0.947	-
Yolov3[18]	Det.	Darknet-53	-	0.363
SSD[15]	Det.	ResNet	-	0.406
MaskRCNN[7]	Seg. & Det.	R50	0.452	0.376
AP-MTL[9]	Seg. & Det.	Trfmr	0.947	0.392
MT3DNet(ours)	Seg. & Det.	Trfmr	0.966	0.451