Salvato in:
Dettagli Bibliografici
Autori principali: Zhao, Shuting, Du, Chenkang, Qi, Kristin, Chen, Xinrong, Di, Xinhan
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:https://arxiv.org/abs/2410.00979
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866913539842637824
author Zhao, Shuting
Du, Chenkang
Qi, Kristin
Chen, Xinrong
Di, Xinhan
author_facet Zhao, Shuting
Du, Chenkang
Qi, Kristin
Chen, Xinrong
Di, Xinhan
contents Adaptation methods are developed to adapt depth foundation models to endoscopic depth estimation recently. However, such approaches typically under-perform training since they limit the parameter search to a low-rank subspace and alter the training dynamics. Therefore, we propose a full-parameter and parameter-efficient learning framework for endoscopic depth estimation. At the first stage, the subspace of attention, convolution and multi-layer perception are adapted simultaneously within different sub-spaces. At the second stage, a memory-efficient optimization is proposed for subspace composition and the performance is further improved in the united sub-space. Initial experiments on the SCARED dataset demonstrate that results at the first stage improves the performance from 10.2% to 4.1% for Sq Rel, Abs Rel, RMSE and RMSE log in the comparison with the state-of-the-art models.
format Preprint
id arxiv_https___arxiv_org_abs_2410_00979
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Towards Full-parameter and Parameter-efficient Self-learning For Endoscopic Camera Depth Estimation
Zhao, Shuting
Du, Chenkang
Qi, Kristin
Chen, Xinrong
Di, Xinhan
Computer Vision and Pattern Recognition
Artificial Intelligence
Adaptation methods are developed to adapt depth foundation models to endoscopic depth estimation recently. However, such approaches typically under-perform training since they limit the parameter search to a low-rank subspace and alter the training dynamics. Therefore, we propose a full-parameter and parameter-efficient learning framework for endoscopic depth estimation. At the first stage, the subspace of attention, convolution and multi-layer perception are adapted simultaneously within different sub-spaces. At the second stage, a memory-efficient optimization is proposed for subspace composition and the performance is further improved in the united sub-space. Initial experiments on the SCARED dataset demonstrate that results at the first stage improves the performance from 10.2% to 4.1% for Sq Rel, Abs Rel, RMSE and RMSE log in the comparison with the state-of-the-art models.
title Towards Full-parameter and Parameter-efficient Self-learning For Endoscopic Camera Depth Estimation
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2410.00979