医学图像分割与分类的基础模型 VISION-MAE - CSDN博客 Here, we present a novel foundation model, VISION- MAE, specifically designed for medical imaging Specifically, VISION-MAE is trained on a dataset of 2 5 million unlabeled images from various modalities (CT, MR, PET, X-rays, and ultrasound), using self-supervised learning techniques
GitHub - ml-jku MAE-CT Each step of MAE-CT requires its own yaml file where the later steps require a reference to a checkpoint of a previous step This can be defined by changing the stage_id of the initializer objects within the yaml
CD-MAE: Contrastive Dual-Masked Autoencoder Pre-Training Model . . . - MDPI Therefore, our proposed model, CD-MAE, is fully pre-trained on unlabeled PCB CT images to obtain an encoder with excellent feature representation capability The encoder can learn deeper features of the image with better robustness
[2304. 10520] Contrastive Tuning: A Little Help to Make Masked . . . To this end, we introduce Masked Autoencoder Contrastive Tuning (MAE-CT), a sequential approach that utilizes the implicit clustering of the Nearest Neighbor Contrastive Learning (NNCLR) objective to induce abstraction in the topmost layers of a pre-trained MAE
Video-CT MAE: Self-supervised Video-CT Domain Adaptation for Vertebral . . . We propose a framework that allows Vision Transformers to efectively detect vertebral fractures in 3D CT images despite a low data regime, outperforming CNN-based methods while providing inherent interpretability through attention visualizations
GFPP-MAE: gradient-guided frequency reconstruction and position . . . To address these challenges, this paper proposes the GFPP-MAE model, which consists of the gradient-guided frequency re-construction module (GFRM), the absolute position prediction module (APPM) and the relative position prediction module (RPPM)