Update import file

convertDLC1TO2.py

@@ -6,28 +6,12 @@
 @author: alex
 """
 
-'''
-Steps:
-    1. copy all labeled data files (folders of images) + CollectedData_user.h5 files to labeled-data
-    
-'''
-
-
-
 from pathlib import Path
 import os, yaml, distutils
 import pandas as pd
 import numpy as np
 from skimage import io
 
-'''
-#First create the project:
-scorer='Mackenzie'
-project='openfield
-pcf=deeplabcut.create_new_project(project,scorer,['/home/alex/Hacking/DLCutils/m3v1mp4.mp4'],copy_videos=True)
-
-'''
-
 def read_config(config):
     """
     Reads config file
@@ -36,48 +20,84 @@ def read_config(config):
         cfg = yaml.load(ymlfile)
     return(cfg)
 
-scorer='Mackenzie'
-project='openfield'
-date='2018-10-30'
+'''
+Attention: this is not completely streamlined, please read carefully. It is assumed that this file 
+is in the basepath (see below), where the new project was created.
+
+Preparation steps:
+
+    1. Create a new project (with DLC2 code)
+    2. Copy all labeled data files (folders of images) into the labeled-data folder
+    3. Copy the CollectedData_scorer.h5 file into the labeled data folder
+    4. (Ideally) add the videos from which these folders were extractd to the videos folder (of the new project)
+    
+Then the following code will chop up the CollectedData_scorer.h5 file and put it in each folder individually. It will
+update the body part list and the video_sets list. Note that it is assumed that the second image per folder is cropped just 
+like the video itself. [The first image is extracted non-cropped in DLC 1].
+
+The code does not carry over trained networks and other things, as this can be quickly recomputed (and is tricky)
+'''
+
+
+# Edit the variables of the new project:
+
+scorer='Alex'
+project='Trailing'
+date='2018-10-05'
+
+# edit the body parts according to the old DLC1 project:
+bodyparts = ["snout","leftear","rightear","tailbase"]
+
 
-#path_config_file=deeplabcut.create_new_project(project,scorer,['/home/alex/Hacking/Social_Behavior/1802202_Day6 Dummy_DeepLabCutlabeled.mp4'],copy_videos=False)
+# set path where the project is:
+basepath='/home/alex/'
 
-basepath='/home/alex/Hacking/DLCutils'
 projectfolder=project+'-'+scorer+'-'+date
+
 path_config_file = os.path.join(projectfolder,'config.yaml')
 bflabeledata=os.path.join(basepath,projectfolder,'labeled-data')
 
+videopath=os.path.join(basepath,projectfolder,'videos')
 
 #Load config file and labeled data
 cfg = read_config(path_config_file)
 DataCombined = pd.read_hdf(os.path.join(bflabeledata,'CollectedData_' + scorer + '.h5'), 'df_with_missing')
 
-# load folders containing images:
+# Load folders containing images:
 folders = [f for f in os.listdir(bflabeledata) if 'Collected' not in f]
 print(folders)
 
-for fn in folders: #rename folders (old DLC1.2 convention)
+# The next step renames folders with labeled frames
+# Thereby, the convection with 'Results' is based on some other DLC version that was not widely circulated
+for fn in folders: 
     if 'Results' in fn:
         try:
             os.rename(os.path.join(bflabeledata,fn),os.path.join(bflabeledata,fn.split('Results')[1]))
         except OSError: #directory not empty!
             distutils.dir_util.copy_tree(os.path.join(bflabeledata,fn),os.path.join(bflabeledata,fn.split('Results')[1]))
             print("Merging", os.path.join(bflabeledata,fn))
-            #NOW DELETE!!!
 
-#%%
 
 folders = [f for f in os.listdir(bflabeledata) if 'Collected' not in f]
+print("The following folders were found:")
 print(folders)
 
 
+scorer=list(set(DataCombined.columns.get_level_values(0)))[0]
+bpts=list(set(DataCombined.columns.get_level_values(1)))
+
+print("Labels and bpts in folder are:", scorer, bpts)
+print("MAKE SURE THEY ARE IDENTICAL TO YOUR PROJECT OR EDIT!")
+
+
+
+# Update the index to the relative path of the image
 for j,fn in enumerate(DataCombined.index):
     if 'Results' in fn:
         DataCombined.index.values[j]=os.path.join('labeled-data',fn.split('Results')[1]) #FULL PATH!
     else:
         DataCombined.index.values[j]=os.path.join('labeled-data',fn) #FULL PATH!for j,fn in enumerate(DataCombined.index):
 
-
 # Extracting data for each folder.
 for folder in folders:
     frames2extract=[]
@@ -91,14 +111,11 @@ def read_config(config):
     DF.to_hdf(os.path.join(bflabeledata,folder, 'CollectedData_' + scorer +".h5"),'df_with_missing',format='table',mode='w')
 
     #adding corresponding video...
-    #ny,nx,ncols=np.shape(io.imread(os.path.join(projectfolder,DF.index[0])))
-    #cfg['video_sets'].update({os.path.join(videopath,folder+'.mp4') : {'crop': ', '.join(map(str, [0, nx, 0, ny]))}})
+    ny,nx,ncols=np.shape(io.imread(os.path.join(projectfolder,DF.index[1])))
+    cfg['video_sets'].update({os.path.join(videopath,folder+'.mp4') : {'crop': ', '.join(map(str, [0, nx, 0, ny]))}})
 
-#%%
-
+
 #Now update list of bodyparts!
-bodyparts = ["snout","leftear","rightear","shoulder"]
-
 cfg['bodyparts']=bodyparts
 
 with open(str(path_config_file), 'w') as ymlfile: