Skip to content

Latest commit

 

History

History
39 lines (33 loc) · 12.9 KB

MODEL_ZOO.md

File metadata and controls

39 lines (33 loc) · 12.9 KB

SKM-TEA Model Zoo and Baselines

Introduction

This file is a running document of baselines and state-of-the-art models for the SKM-TEA v1 dataset.

Reconstruction Baselines

Name Input Train Acc. Description Download
Unrolled E1 (6x) E1 6x 8 resblocks,
256 channels each
🤗 | cfg | model
Unrolled E2 (6x) E2 6x 8 resblocks,
256 channels each
🤗 | cfg | model
Unrolled E1+E2 (6x) E1+E2 6x 8 resblocks,
256 channels each
🤗 | cfg | model
Unrolled E1$\oplus$E2 (6x) E1$\oplus$E2 6x 8 resblocks,
256 channels each
🤗 | cfg | model
U-Net E1 (6x) E1 6x 4 pooling, 32^p, leaky-relu activation 🤗 | cfg | model
U-Net E2 (6x) E2 6x 4 pooling, 32^p, leaky-relu activation 🤗 | cfg | model
U-Net E1+E2 (6x) E1+E2 6x 4 pooling, 32^p, leaky-relu activation 🤗 | cfg | model
U-Net E1$\oplus$E2 (6x) E1$\oplus$E2 6x 4 pooling, 32^p, leaky-relu activation 🤗 | cfg | model
Unrolled E1 (8x) E1 8x 8 resblocks,
256 channels each
🤗 | cfg | model
Unrolled E2 (8x) E2 8x 8 resblocks,
256 channels each
🤗 | cfg | model
Unrolled E1+E2 (8x) E1+E2 8x 8 resblocks,
256 channels each
🤗 | cfg | model
Unrolled E1$\oplus$E2 (8x) E1$\oplus$E2 8x 8 resblocks,
256 channels each
🤗 | cfg | model
U-Net E1 (8x) E1 8x 4 pooling, 32^p, leaky-relu activation 🤗 | cfg | model
U-Net E2 (8x) E2 8x 4 pooling, 32^p, leaky-relu activation 🤗 | cfg | model
U-Net E1+E2 (8x) E1+E2 8x 4 pooling, 32^p, leaky-relu activation 🤗 | cfg | model
U-Net E1$\oplus$E2 (8x) E1$\oplus$E2 8x 4 pooling, 32^p, leaky-relu activation 🤗 | cfg | model

Segmentation Baselines

Name Input Description Download
V-Net E1 E1 Monai v0.5.3 🤗 | cfg | model
V-Net E2 E2 Monai v0.5.3 🤗 | cfg | model
V-Net E1$\oplus$E2 E1$\oplus$E2 Monai v0.5.3 🤗 | cfg | model
V-Net RSS RSS Monai v0.5.3 🤗 | cfg | model
U-Net E1 E1 5 pooling layers, 32^p channels 🤗 | cfg | model
U-Net E2 E2 5 pooling layers, 32^p channels 🤗 | cfg | model
U-Net E1$\oplus$E2 E1$\oplus$E2 5 pooling layers, 32^p channels 🤗 | cfg | model
U-Net RSS RSS 5 pooling layers, 32^p channels 🤗 | cfg | model