| → | → | → | |||||
| Normalize & downsample images | Pick training particles, or analyze picks | Train a neural network | Extract particle coordinates |
Citation: Bepler, T., Morin, A., Rapp, M., Brasch, J., Shapiro, L., Noble, A.J., Berger, B. (2019). Positive-unlabeled convolutional neural networks for particle picking in cryo-electron micrographs. Nature Methods
Topaz Image Pre-processing Command Generator |
Create a command for pre-processing images (downsampling and normalization) for use in this Topaz GUI and in Topaz training.
Parameters:
topaz preprocess /path/to/input/images/*.mrc --scale 4 --sample 1 --num-workers -1 --format mrc,png --device 0 --niters 100 --alpha 900 --beta 1 --verbose --destdir /path/to/output/preprocessed/images/
"./" indicates that the folder containing the topaz.html application file also contains the images in this project. For example: /datasets/VOC2012/JPEGImages/ or C:\Documents\data\ (note the trailing / and \)Filename:
We recommend that you update the default path in to the folder which contains this image.
A temporary fix is to use to manually locate and add this file. We do not recommend this approach because it requires you to repeat this process every time your load this project in the VIA application.
Topaz is a pipeline for particle detection in cryo-electron microscopy images using convolutional neural networks trained from positive and unlabeled examples. The original research can be found here. Topaz is released under the GPLv3 license.
Topaz GUI uses VGG Image Annotator (VIA), which is an image annotation tool that can be used to define regions in an image and create textual descriptions of those regions. VIA is an open source project developed at the Visual Geometry Group and released under the BSD-2 clause license.
Here is a list of some salient features of the Topaz GUI version of VIA:
For more Topaz details, visit http://cb.csail.mit.edu/cb/topaz/.
For more VIA details, visit http://www.robots.ox.ac.uk/~vgg/software/via/.
VIA is Copyright © 2016-2019, Abhishek Dutta,Visual Geometry Group, Oxford University and VIA Contributors.
Pick or Analyze Particles |
Pick particles for training networks or analyze Topaz picks. Follow the steps below:
1) To start picking, or add images from
2) Pick 100-1,000 particles (more is better) representing all orientations.
• Pick across several representative images.
• Not all particles need to be picked from each image.
3) Export your particle picks and proceed to Topaz training.
• and regularly. You may later to continue.
• For further help, see the page.
• your Topaz particle picks as a CSV file and review the picks. ◦ Press L on your keyboard to show the particle number, then the Up ↑ key will toggle the Topaz particle scores. Zoom-in on the image with + to inspect. Or • Load your micrographs and particles in .star format into Relion. Picks may be colored based on Topaz score by using the AutopickFigureOfMerit column in the .star file.
Topaz Training Command Generator |
Create a command for training a network using particle picks.
Parameters:
topaz train --train-images /path/to/preprocessed/images/ --train-targets /path/to/training_particles.csv --k-fold 5 --fold 0 --radius 3 --model resnet8 --image-ext .mrc --units 32 --dropout 0.0 --bn on --unit-scaling 2 --ngf 32 --method GE-binomial --autoencoder 0 --num-particles 300 --l2 0 --learning-rate 0.0002 --minibatch-size 256 --minibatch-balance 0.0625 --epoch-size 5000 --num-epochs 10 --num-workers -1 --test-batch-size 1 --device 0 --save-prefix /output/path/model --output /output/path/results.txt
Topaz Picks Extraction Command Generator |
Create commands for extracting and converting particle coordinates.
Parameters:
topaz extract /path/to/preprocessed/images/*.mrc --model /path/to/model_epoch##.sav --radius 8 --threshold -6 --up-scale 1 --batch-size 1 --min-radius 5 --max-radius 100 --step-radius 5 --num-workers -1 --device 0 --output /path/to/extracted/particles.txt
topaz convert /path/to/extracted/particles.txt --verbose 1 --output /path/to/extracted/particles.star
topaz convert /path/to/extracted/particles.txt --verbose 1 --output /path/to/extracted/particles.csv
More Tools |
Create a command for denoising micrographs using an implementation of Noise2Noise.
Create a command for denoising a particle stack using an implementation of Noise2Noise.
Create a command for converting between Topaz GUI CSV files, STAR files, or Topaz txt files.
Create a command for extracting a particle stack from micrographs using a Topaz .txt coordinate file.
Parameters:
Advanced options
Command:
topaz denoise /path/to/input/images/*.mrc --model L2 --device 0 --format mrc --bin 1 --patch-size 1536 --patch-padding 384 --normalize --output /path/to/output/folder
Denoise Particle Stack
Parameters:
Command:
topaz denoise /path/to/particle/stack.mrcs --model L2 --device 0 --format mrc --bin 1 --stack --normalize --output /path/to/denoised/stack.mrcs
Convert | Threshold Picks
Parameters:
Specific filetype options
Command:
topaz convert /path/to/input/coordinate/file.[csv|star|txt] --threshold -9999 --up-scale 1 --down-scale 1 --voltage -1 --detector-pixel-size -1 --magnification -1 --amplitude-contrast -1 --boxsize 0 --image-ext .mrc --verbose 1 --output /path/to/output/coordinate/file.[csv|star|txt]
Extract Particle Stack
Parameters:
Command:
topaz particle_stack /path/to/Topaz/coordinate/file.txt --image-root /path/to/full/micrographs/ --size 192 --threshold -9999 --resize -1 --image-ext .mrc --output /path/to/output/particles.mrcs