Differences

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

--- janni_tutorial [2019/09/16 09:36]
twagner [Configuration]
+++ janni_tutorial [2019/09/16 09:42]
twagner [Training a model for your data]
@@ Line 45: / Line 45: @@
 Press "start" and the config will be written in the specified //config_out_path//.
- In the following I assume you named it //config.json//. It should look like this:
+If you would like to use the command line, you can get a descriptio of all parameters with:
-<code json config.json>
+<code>
-{
+janni_denoise.py config -h
-    "model" : {
+</code>
-        "architecture":         "unet",
-        "patch_size":           1024
-    },
-    "train": {
+The following command will create the same config file as with the GUI:
-        "movie_dir":            "/path/to/movie/directory/",
+<code>
-        "even_dir":             "even_averages/",
+janni_denoise.py config ~/example/config_janni.json --movie_dir ~/example/movies/ --even_dir ~/example/even/ --odd_dir ~/example/odd/
-        "odd_dir":              "odd_averages/",
-        "batch_size":           4,
-        "learning_rate":        1e-3,
-        "nb_epoch":             100,
-        "saved_weights_name":   "mymodel.h5"
-    }
-}
 </code>
-The fields have the following meaning:
-  * **architecture**: CNN Architecture. Right now, only unet is supported.
-  * **patch_size**: The image will be denoised in patches. This field defines the patch size.
-  * **movie_dir**: Path to the directory with the movie files. If an average exists already in even_dir or odd_dir (see below), it will be skipped.
-  * **even_dir**: For each movie in movie_dir, an average based on the //even// frames is calculated and saved in even_dir.
-  * **odd_dir**: For each movie in movie_dir, an average based on the //odd// frames is calculated and saved in odd_dir.
-  * **batch_size**: How many patches are in one mini-batch. If you have memory problems (e.g with cards < 8GB memory), you can try to reduce this value.
-  * **learning_rate**: Learning rate, should not be changed.
-  * **nb_epoch**: Number of epochs to train. More epochs seems to only slightly improve the results.
-  * **saved_weights_name**: Filename of your model.
-In principle you only have to adapt the paths. The other could keep as they are.
 We typically use at least 30 movies (unaligned) to train the model. Less might also work, more work often much better.

User Tools

Site Tools

Differences

Page Tools