User Tools

Site Tools

cryolo_release_note_110

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revision Previous revision
Next revision 		Previous revision
Last revision Both sides next revision
cryolo_release_note_110 [2018/08/23 11:17]
twagner 		cryolo_release_note_110 [2018/08/23 13:28]
twagner
Line 1: Line 1:
 ====== crYOLO 1.1.0 release notes ====== ====== crYOLO 1.1.0 release notes ======
-===== Changes ===== +===== Changes crYOLO ===== 
-  * crYOLO finally now supports Filaments (more here)+  * crYOLO now supports filaments (more here)
   * New evaluation tool (more here)   * New evaluation tool (more here)
-  * Support empty box files for training on particle-free images +  * Supports empty box files for training on particle-free images 
-  * Extended data augmentation: Horizontal flip and flip along both axis. +  * Extended data augmentation: Horizontal flip and flip along both axes 
-  * Experimental support of periodic restarts during training (with --warm_restarts).+  * Experimental support of periodic restarts during training (with --warm_restarts) 
 + 
 +===== Changes crYOLO Boxmanager ===== 
 +  * Support of visualization of EMAN2 helical coordinates (particle coordinates) 
 +  * New boxes could be loaded with a new color while keeping the old. 
 +  * Several bug fixes 
  
  
 ===== Picking filaments - Using a model trained for your data ===== ===== Picking filaments - Using a model trained for your data =====
 +Since version 1.1.0 crYOLO supports picking filaments. 
  
 ==== Data preparation ==== ==== Data preparation ====
-As described [[pipeline:window:cryolo#data_preparation|above]], filtering your image using a low-pass filter is probably a good idea. +As described [[pipeline:window:cryolo#data_preparation|previously]], filtering your image using a low-pass filter is probably a good idea. 
  
 After this is done, you have to prepare training data for your model. After this is done, you have to prepare training data for your model.
Line 19: Line 26:
 </code> </code>
  
-After tracing your training data in e2helixboxer, export them using //File -> Save//. Make sure that you export particle coordinates as this the only format supported right now (see screenshot). In the following it is expected that you exported into a folder called "train_annot".+After tracing your training data in e2helixboxer, export them using //File -> Save//. Make sure that you export particle coordinates as this the only format supported right now (see screenshot). In the following example, it is expected that you exported into a folder called "train_annot".
  
 ==== Configuration ==== ==== Configuration ====
-You can configure it the same way as for a "normal" project. We recommend to use the [[pipeline:window:cryolo#alternative_configurationyolo_network_in_patch_mode_-_beta|patch mode]].+You can configure it the same way as for a "normal" project. We recommend to use [[pipeline:window:cryolo#alternative_configurationyolo_network_in_patch_mode_-_beta|patch mode]]. 
 ==== Training ==== ==== Training ====
  
-In principle, there is not much difference in training crYOLO for filament picking and particle picking. However, to our experience the warm-up phase and training need a little bit more time:+In principle, there is not much difference in training crYOLO for filament picking and particle picking. For project with roughly 20 filaments per image we successfully trained on 40 images (=> 800 filaments). However, in our experience the warm-up phase and training need a little bit more time:
  
 **1. Warm up your network** **1. Warm up your network**
Line 45: Line 53:
  
   * //- -filament//: Option that tells crYOLO that you want to predict filaments   * //- -filament//: Option that tells crYOLO that you want to predict filaments
-  * //-fw//: Filament width (pixel) +  * //-fw//: Filament width (pixels) 
-  * //-bd//: Inter-Box distance (pixel).+  * //-bd//: Inter-Box distance (pixels).
  
 Let's assume you want to pick a filament with a width of 100 pixels. The box size is 200x200 and you want a 90% overlap, than the picking command would be: Let's assume you want to pick a filament with a width of 100 pixels. The box size is 200x200 and you want a 90% overlap, than the picking command would be:
Line 54: Line 62:
  
 The directory boxes will be created and all results are saved there. The format is the eman2 helix format with particle coordinates. The directory boxes will be created and all results are saved there. The format is the eman2 helix format with particle coordinates.
 +
 ==== Visualize the results ==== ==== Visualize the results ====
-You can use the boxmanager as described [[pipeline:window:cryolo#visualize_the_results|above]].+You can use the boxmanager as described [[pipeline:window:cryolo#visualize_the_results|previously]].
  
-===== New evaluation tool =====+===== Evaluate your results =====
  
 The new evaluation tool allows you, based on your validation data, to get statistics about your training.  The new evaluation tool allows you, based on your validation data, to get statistics about your training. 
-If you followed the tutorial, the validation data is selected randomly. With crYOLO 1.1.0 a run file for each training is created and saved into the folder runfiles/ in your project directory. This run file contains, which files were selected for validation and you can run your evaluation as follows:+If you followed the tutorial, the validation data are selected randomly. With crYOLO 1.1.0 a run file for each training is created and saved into the folder runfiles/ in your project directory. This run file contains which files were selected for validation, and you can run your evaluation as follows:
 <code> <code>
 cryolo_evaluation.py -c config.json -w model.h5 -r runfiles/run_20180821-144617.json  cryolo_evaluation.py -c config.json -w model.h5 -r runfiles/run_20180821-144617.json 
Line 68: Line 77:
 {{ :eval_tool.png?900 |}} {{ :eval_tool.png?900 |}}
  
-The table contains several statics:+The table contains several statistics:
   * AUC: Area under curve of the precision-recall curve. Overall summary statistics. Perfect classifier = 1, Worst classifier = 0   * AUC: Area under curve of the precision-recall curve. Overall summary statistics. Perfect classifier = 1, Worst classifier = 0
-  * Topt: Optimal confidence threshold with respect to the F1 score. It might not be ideal for your picking, as the F1 score weights recall and precision the same. However, in SPA the Recall is often more important than the precision.  +  * Topt: Optimal confidence threshold with respect to the F1 score. It might not be ideal for your picking, as the F1 score weighs recall and precision equally. However in SPA, recall is often more important than the precision.  
   * R (Topt): Recall using the optimal confidence threshold.    * R (Topt): Recall using the optimal confidence threshold. 
   * R (0.3): Recall using a confidence threshold of 0.3.   * R (0.3): Recall using a confidence threshold of 0.3.
Line 80: Line 89:
   * F1 (0.3): Harmonic mean of precision and recall using a confidence threshold of 0.3.   * F1 (0.3): Harmonic mean of precision and recall using a confidence threshold of 0.3.
   * F1 (0.2): Harmonic mean of precision and recall using a confidence threshold of 0.2.   * F1 (0.2): Harmonic mean of precision and recall using a confidence threshold of 0.2.
-  * IOU (Topt): Intersection over union of the auto-picked particle and the corresponding ground-truth box. As higher, the better. Evaluated with the optimal confidence threshold. +  * IOU (Topt): Intersection over union of the auto-picked particles and the corresponding ground-truth boxes. The higher, the better -- evaluated with the optimal confidence threshold. 
-  * IOU (0.3): Intersection over union of the auto-picked particle and the corresponding ground-truth box. As higher, the better. Evaluated with a confidence threshold of 0.3. +  * IOU (0.3): Intersection over union of the auto-picked particles and the corresponding ground-truth boxes. The higher, the better -- evaluated with a confidence threshold of 0.3. 
-  * IOU (0.2): Intersection over union of the auto-picked particle and the corresponding ground-truth box. As higher, the better. Evaluated with a confidence threshold of 0.2.+  * IOU (0.2): Intersection over union of the auto-picked particles and the corresponding ground-truth boxes. The higher, the better -- evaluated with a confidence threshold of 0.2.
  
 If the training data consists of multiple folders, then evaluation will be done for each folder separately.  If the training data consists of multiple folders, then evaluation will be done for each folder separately. 
cryolo_release_note_110.txt · Last modified: 2018/08/24 08:19 by twagner

Page Tools

Except where otherwise noted, content on this wiki is licensed under the following license: GNU Free Documentation License 1.3
GNU Free Documentation License 1.3 Donate Powered by PHP Valid HTML5 Valid CSS Driven by DokuWiki