search

tbepler / topaz

Pipeline for particle picking in cryo-electron microscopy images using convolutional neural networks trained from positive and unlabeled examples. Also featuring micrograph and tomogram denoising with DNNs.
GNU General Public License v3.0
169 stars 63 forks source link

Is there a way to decrease the restriction distance between predicted particles? #62

Closed Zivvvo closed 4 years ago

Zivvvo commented 4 years ago

Greetings, I am a researcher working on Cryo-EM 3D reconstruction of microtubule proteins, specifically the doublet microtubules from cilia and flagella. We are currently using topaz to perform automatic picking of filamentous structures. Currently, topaz is picking very well except that there seems to be a fixed minimum distance between neighboring particles. How to achieve a denser extraction and which of the parameters should be changed during the extraction?

Extraction command: topaz extract -r 40 -m saved_models_full_pick/model_epoch10.sav \ -o data/topaz_full_pick/predicted_particles_all.txt \ data/processed/training_micrographs/*.mrc

Picking results: image

Ground truth: image

tbepler commented 4 years ago

I'm glad you're getting good results! Increasing the extracted particle density is as easy as reducing the radius parameter for topaz extract. This parameter sets the minimum distance (in pixels) allowed between particle centers. For example, if you wanted to allow 5 pixels between centers, you would replace -r40 with -r5 in your command.

tbepler commented 4 years ago

Closing this issue since the questions seems to be resolved!