Update

our_scripts/config_bayes.yml

@@ -1,12 +1,11 @@
 bayes:
-  iterations: 2
+  iterations: 10
 train:
   epochs:
     type: discrete
-    values: [10]
+    values: [30]
   batch-size:
     type: discrete
-    #values: [128]
     min: 1
     max: 5
     step: 1
@@ -18,12 +17,12 @@ train:
   img-size-start:
     type: discrete
     min: 512
-    max: 576
+    max: 1088
     step: 64
   img-size-end:
     type: discrete
     min: 512
-    max: 576
+    max: 1088
     step: 64
   rect:
     type: discrete
@@ -122,16 +121,16 @@ detect:
   test-img-size:
     type: discrete
     min: 512
-    max: 576
+    max: 1088
     step: 64
   conf-thres:
     type: continuous
-    min: 0.0
+    min: 0.3
-    max: 1.0
+    max: 0.6
   iou-thres:
     type: continuous
-    min: 0.0
+    min: 0.3
-    max: 1.0
+    max: 0.6
   classes:
   agnostic-nms:
   augment:

our_scripts/run_yolov3_process_bayes.py

@@ -8,7 +8,8 @@ import traceback
 import GPyOpt
 from config_bayes import Configuration
 
-from utils import call_subprocess, get_values_from_conff_matrix
+from utils import call_subprocess, get_values_from_conff_matrix, load_previous_bayes_experiments
+
 
 dir_path = os.path.dirname(os.path.realpath(__file__))
 PROJECT_ROOT = os.path.join(dir_path, '..')
@@ -19,7 +20,6 @@ config = Configuration(bayes_config_yaml)
 date_string = datetime.datetime.now().strftime('%Y-%m-%d_%H-%M-%S')
 bayes_params_file = open(os.path.join(PROJECT_ROOT, config.experiments.dir, f"{date_string}_bayes_params.txt"), 'a+')
 
-
 def call_training_script(gaussian_hyps):
     cmd = 'python -u train.py'
     cmd += ' --epochs ' + gaussian_hyps['epochs'].__str__()
@@ -149,7 +149,7 @@ def yolov3(x):
         y_dict = get_values_from_conff_matrix(conf_matrix_path)
 
         # tutaj wzór na wyliczanie funkcji 
-        y_val = 1 - ((y_dict['match'] * 10 - y_dict['false positives'] * 3) / y_dict['mistakes'])
+        y_val = 1 - (y_dict['match'] * 10 - y_dict['false positives'] * 3 - y_dict['mistakes']) / y_dict['all labels']
 
         # zapisywanie do pliku zadeklarowanego globalnie
         line = "\t".join([bayes_hyps.__str__(), str(y_val)])
@@ -168,14 +168,9 @@ def yolov3(x):
 if __name__ == '__main__':
     bounds = config.get_bayes_bounds()
 
-    # for b in bounds:
-    #   print(b)
-
-    # tutaj będzie wczytywanie z poprzednich eksperymentów plik bayes_params
-    X = None
-    Y = None
-
     os.chdir(PROJECT_ROOT)  # change to project root directory
+    # wczytywanie z poprzednich eksperymentów plik bayes_params
+    X, Y =  load_previous_bayes_experiments(config.experiments.dir)
 
     bayes_optimizer = GPyOpt.methods.BayesianOptimization(f=yolov3, domain=bounds, X=X, Y=Y, verbosity=True,
                                                           initial_design_numdata=2)

our_scripts/utils.py

@@ -1,5 +1,9 @@
+import ast
 import io
+import os
 import subprocess
+import numpy as np
+from glob import glob
 
 
 def call_subprocess(cmd):
@@ -53,3 +57,37 @@ def get_bayes_params_as_dict(x):
         'conf-thres': float(x[:, 26]),
         'iou-thres': float(x[:, 27])
     }
+
+
+def load_previous_bayes_experiments(experiments_dir):
+    paths = list(glob(os.path.join(experiments_dir, '*bayes_params.txt')))
+    y_values = []
+    x_values = []
+
+    for p in paths:
+        file = open(p, 'r')
+        lines = file.readlines()
+        for line in lines:
+            try:
+                dict_str, y_val = line.split('\t')
+                bayes_dict = ast.literal_eval(dict_str)
+                bayes_values = dict_to_numpy(bayes_dict)
+                x_values.append(bayes_values)
+                y_values.append(float(y_val))
+            except:
+                raise Exception(f"Cannot parse line {line} from file {p}")
+    return np.array(x_values), np.array(y_values).reshape((len(y_values), 1))
+
+
+def dict_to_numpy(d):
+    x = []
+    for key, value in d.items():
+        if type(value) == bool:
+            x.append(int(value))
+        elif key == 'img-size':
+            start_img_size, end_img_size = value.split(" ")
+            x.append(int(start_img_size))
+            x.append(int(end_img_size))
+        else:
+            x.append(float(value))
+    return x