text stringclasses 4
values |
|---|
image0000 |
image0600 |
SEM-Imaging_TLD-BSE_q1_x01_y01_s0053.filtered0200 |
SEM-Imaging_TLD-BSE_q1_x01_y01_s0053.filtered1328 |
YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
This dataset contains two vEM datasets used for 3D reconstruction in the EMCFsys paper. The AP and IE directories store the training images, corresponding labels, and dataset splits, respectively. The Volume folder contains the 2D image sequences of vEM.
In EMCFsys paper fig6, low resolution images were first made super-resolution to High-resolution. And then train segmentation model. At last make 3D reconstruction from 2D-slice segmentation inference.
For the anterior pituitary dataset, the raw low-resolution volume (1024×884×854) was acquired via FIB-SEM. We then performed super-resolution upsampling using EMCellFiner to obtain the high-resolution volume of 4096×3536×854.
For the intestinal epithelial dataset, the raw low-resolution volume (1750×1500×2251) was acquired via FIB-SEM. Similarly, super-resolution upsampling was performed with EMCellFiner, yielding a high-resolution volume of 7000×6000×2251.
Subsequently, we selected 1–4 slice images from each volume to fine-tune the EMCFound model for organelle segmentation on 2D slices. All segmentation outputs can be visualized as 3D structures using Amira or Arivis Pro.
license: apache-2.0
- Downloads last month
- 220