You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
PartiNet is a three-stage pipeline for automated particle picking in cryo-EM micrographs, combining advanced denoising with state-of-the-art deep learning detection.
4
6
@@ -9,12 +11,12 @@ Use our pretrained model at [Model Weights](https://huggingface.co/MihinP/PartiN
9
11
10
12
## Features
11
13
12
-
-🧹 Heuristic denoising for improved signal-to-noise ratio
13
-
-🎯 Dynamic deep learning particle detection
14
-
-⚡ Multi-GPU support for faster processing
15
-
-🔄 Seamless integration with cryoSPARC and RELION workflows
16
-
-📊 Confidence-based particle filtering
17
-
-🖼️ Visual detection validation
14
+
- Heuristic denoising for improved signal-to-noise ratio
15
+
- Dynamic deep learning particle detection
16
+
- Multi-GPU support for faster processing
17
+
- Seamless integration with cryoSPARC and RELION workflows
18
+
- Confidence-based particle filtering
19
+
- Visual detection validation
18
20
19
21
## Prerequisites
20
22
@@ -49,12 +51,12 @@ docker run --gpus all -v /path/to/your/data:/data \
49
51
50
52
```
51
53
project_directory/
52
-
├── motion_corrected/ # 📁 Input micrographs
53
-
├── denoised/ # 🧹 Denoised outputs
54
-
├── exp/ # 🎯 Detection results
55
-
│ ├── labels/ # 📋 Coordinates
56
-
│ └── ... # 🖼️ Visualizations
57
-
└── partinet_particles.star # ⭐ Final output
54
+
├── motion_corrected/ # Input micrographs
55
+
├── denoised/ # Denoised outputs
56
+
├── exp/ # Detection results
57
+
│ ├── labels/ # Coordinates
58
+
│ └── ... # Visualizations
59
+
└── partinet_particles.star # Final output
58
60
```
59
61
60
62
## Pipeline Stages
@@ -94,7 +96,8 @@ partinet star \
94
96
-`*.png`: Visualization overlays
95
97
96
98
3.**STAR File** (`partinet_particles.star`)
97
-
- Ready for RELION processing
99
+
- Ready for CryoSPARC processing
100
+
- Additional commands available for RELION compatible output
98
101
99
102
## Advanced Usage
100
103
@@ -145,3 +148,5 @@ Citation information will be added upon publication
145
148
For issues and questions:
146
149
- Open an [Issue](https://github.com/WEHI-ResearchComputing/PartiNet/issues)
By default PartiNet outputs denoised images in `png` format. This is necessary for compatibility with the detection architecture. `png` is a lossless compression, however micrographs are normalised from 32 bit depth `mrc` files to 8 bit `png`. `jpg` is also available (eg for making figures) but is not recommended for use due to lossy compression.
169
+
By default PartiNet outputs denoised images in `png` format. This is necessary for compatibility with the detection architecture. `png` is a lossless compression, however micrographs are normalised from 32 bit depth `mrc` files to 8 bit `png`. `jpg` is also available (as a legacy format) but is not recommended for use due to lossy compression.
170
170
171
171
```shell
172
172
# JPEG format (smaller file size, lossy compression)
@@ -96,8 +116,10 @@ After generating your STAR file, you can:
96
116
97
117
:::warning
98
118
99
-
PartiNet Denoise flips micrographs in the y-axis due to matrix transpose operations. If you denoised micrographs in PartiNet ensure that when you import and extract particle coordinates that you toggle `Flip in y` in CryoSPARC or similar in RELION prior to particle extraction.
119
+
If you are using PartiNet v1.0.0, the Denoise flips micrographs in the y-axis due to matrix transpose operations. If you denoised micrographs in PartiNet ensure that when you import and extract particle coordinates that you toggle `Flip in y` in CryoSPARC prior to particle extraction. This has been fixed in v1.0.1+. You can check your version with `partinet --help`
@@ -33,8 +39,8 @@ export default function Home(): ReactNode {
33
39
const{siteConfig}=useDocusaurusContext();
34
40
return(
35
41
<Layout
36
-
title={`Hello from ${siteConfig.title}`}
37
-
description="Description will go into a meta tag in <head />">
42
+
title={`PartiNet: high performance particle picking for cryo-EM`}
43
+
description="PartiNet is a high performance particle picking pipeline for cryo-electron microscopy (cryo-EM) that provides improved identification of rare particle views and up to 7× faster inference than existing tools.">
0 commit comments