相关实验视频

Updated: May 1, 2026

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography

Published on: November 30, 2022

M3AE-蒸:用于医疗视觉语言下游任务的高效蒸模型

Xudong Liang¹, Jiang Xie¹, Mengfei Zhang¹

¹School of Computer Engineering and Science, Shanghai University, Shanghai 200444, China.

Bioengineering (Basel, Switzerland)

|July 29, 2025

概括

本研究介绍了M3AE-Distill,这是一款轻量级的医疗视觉语言模型,使用知识蒸进行高效的预培训. 它实现了与较大的模型可比的高性能,同时显著加快了推断和微调的速度.

科学领域:

人工智能的人工智能
医疗信息学医疗信息学
计算机视觉计算机视觉

背景情况:

多模态蒙面自动编码器 (M3AE) 是强大的医学视觉语言 (VL) 模型,但由于参数大小,阻碍了临床部署.
知识蒸 (KD) 有效地压缩了单调模型,但它对预培训医疗VL骨干模型的应用尚未得到充分确立.

研究的目的:

开发M3AE-Distill,一个轻量级的医疗VL模型,通过知识蒸提高效率,同时保持高性能.
在现实世界临床环境中解决大型M3AE模型的部署挑战.

主要方法:

采用隐藏状态和注意力图蒸来指导学生模型在预培训期间.
引入了以注意力为导向的掩盖策略,以改善细粒度图像-文本对齐.
开发了两个学生变体,M3AE-Distill-Small和M3AE-Distill-Base,用于灵活的准确性-效率权衡.

主要成果:

M3AE-Distill在五个医疗VL数据集和三个任务中展示了有效性.
M3AE-Distill-Base实现了与教师模型相提并论的性能.
实现了显著的加速度:推断时2.11×,微调时2.61×.

结论:

M3AE-Distill为医疗视觉语言任务提供了一种高效和高性能解决方案.
拟议的蒸和掩盖策略可以提高模型的效率,而不会牺牲准确性.
模型变体为部署高效的医疗VL系统提供了实际的选择.

关键词:

深度学习是一种深度学习.知识的蒸知识的蒸.预先培训的培训前培训视觉语言

更多相关视频

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness

Published on: December 6, 2024

Development of a Gaze-Contingent Display Framework Designed for Perceptual and Oculomotor Research with Simulated Central Vision Loss

Development of a Gaze-Contingent Display Framework Designed for Perceptual and Oculomotor Research with Simulated Central Vision Loss

Published on: April 11, 2025

相关概念视频

Vision

Vision

Vision is the result of light being detected and transduced into neural signals by the retina of the eye. This information is then further analyzed and interpreted by the brain. First, light enters the front of the eye and is focused by the cornea and lens onto the retina—a thin sheet of neural tissue lining the back of the eye. Because of refraction through the convex lens of the eye, images are projected onto the retina upside-down and reversed.