Merge pull request OlafenwaMoses#336 from rola93/fix_file_paths

Fix file paths

imageai/Detection/Custom/__init__.py

@@ -110,24 +110,20 @@ def setDataDirectory(self, data_directory):
         :return:
         """
 
-        self.__train_images_folder = os.path.join(data_directory, "train/images/")
-        self.__train_annotations_folder = os.path.join(data_directory, "train/annotations/")
-        self.__validation_images_folder = os.path.join(data_directory, "validation/images/")
-        self.__validation_annotations_folder = os.path.join(data_directory, "validation/annotations/")
+        self.__train_images_folder = os.path.join(data_directory, "train", "images")
+        self.__train_annotations_folder = os.path.join(data_directory, "train", "annotations")
+        self.__validation_images_folder = os.path.join(data_directory, "validation", "images")
+        self.__validation_annotations_folder = os.path.join(data_directory, "validation", "annotations")
 
-        if os.path.exists(os.path.join(data_directory, "cache")) == False:
-            os.makedirs(os.path.join(data_directory, "cache"))
+        os.makedirs(os.path.join(data_directory, "cache"), exist_ok=True)
         self.__train_cache_file = os.path.join(data_directory, "cache", "detection_train_data.pkl")
         self.__validation_cache_file = os.path.join(data_directory, "cache", "detection_test_data.pkl")
 
-        if os.path.exists(os.path.join(data_directory, "models")) == False:
-            os.makedirs(os.path.join(data_directory, "models"))
+        os.makedirs(os.path.join(data_directory, "models"), exist_ok=True)
 
-        if os.path.exists(os.path.join(data_directory, "json")) == False:
-            os.makedirs(os.path.join(data_directory, "json"))
+        os.makedirs(os.path.join(data_directory, "json"), exist_ok=True)
 
-        if os.path.exists(os.path.join(data_directory, "logs")) == False:
-            os.makedirs(os.path.join(data_directory, "logs"))
+        os.makedirs(os.path.join(data_directory, "logs"), exist_ok=True)
 
         self.__model_directory = os.path.join(data_directory, "models")
         self.__train_weights_name = os.path.join(self.__model_directory, "detection_model-")
@@ -152,7 +148,7 @@ def setGpuUsage(self, train_gpus):
         # let it as a string separated by commas
         self.__train_gpus = ','.join([str(gpu) for gpu in train_gpus])
 
-    def setTrainConfig(self,  object_names_array, batch_size= 4, num_experiments=100, train_from_pretrained_model=""):
+    def setTrainConfig(self,  object_names_array, batch_size=4, num_experiments=100, train_from_pretrained_model=""):
 
         """
 
@@ -174,21 +170,17 @@ def setTrainConfig(self,  object_names_array, batch_size= 4, num_experiments=100
                                                                           self.__train_images_folder,
                                                                           self.__train_cache_file, self.__model_labels)
 
-
-
         self.__model_labels = sorted(object_names_array)
         self.__num_objects = len(object_names_array)
 
         self.__train_batch_size = batch_size
         self.__train_epochs = num_experiments
         self.__pre_trained_model = train_from_pretrained_model
 
-
-        json_data = {}
+        json_data = dict()
         json_data["labels"] = self.__model_labels
         json_data["anchors"] = self.__inference_anchors
 
-
         with open(os.path.join(self.__json_directory, "detection_config.json"), "w+") as json_file:
             json.dump(json_data, json_file, indent=4, separators=(",", " : "),
                       ensure_ascii=True)
@@ -219,7 +211,7 @@ def trainModel(self):
             self.__model_labels
 
         )
-        if(self.__training_mode):
+        if self.__training_mode:
             print('Training on: \t' + str(labels) + '')
             print("Training with Batch Size: ", self.__train_batch_size)
             print("Number of Experiments: ", self.__train_epochs)
@@ -287,7 +279,6 @@ def trainModel(self):
         ###############################
         callbacks = self._create_callbacks(self.__train_weights_name, infer_model)
 
-
         train_model.fit_generator(
             generator=train_generator,
             steps_per_epoch=len(train_generator) * self.__train_times,

imageai/Detection/Custom/evaluate.py

@@ -2,16 +2,13 @@
 
 import argparse
 import os
-import numpy as np
 import json
 from imageai.Detection.Custom.voc import parse_voc_annotation
-from imageai.Detection.Custom.yolo import create_yolov3_model
 from imageai.Detection.Custom.generator import BatchGenerator
 from imageai.Detection.Custom.utils.utils import normalize, evaluate
-from keras.callbacks import EarlyStopping, ModelCheckpoint
-from keras.optimizers import Adam
 from keras.models import load_model
 
+
 def _main_(args):
     config_path = args.conf
 
@@ -60,6 +57,7 @@ def _main_(args):
         print(labels[label] + ': {:.4f}'.format(average_precision))
     print('mAP: {:.4f}'.format(sum(average_precisions.values()) / len(average_precisions)))           
 
+
 if __name__ == '__main__':
     argparser = argparse.ArgumentParser(description='Evaluate YOLO_v3 model on any dataset')
     argparser.add_argument('-c', '--conf', help='path to configuration file')    

imageai/Detection/Custom/gen_anchors.py

@@ -1,9 +1,8 @@
 import random
-import argparse
 import numpy as np
 
 from imageai.Detection.Custom.voc import parse_voc_annotation
-import json
+
 
 def IOU(ann, centroids):
     w, h = ann
@@ -24,6 +23,7 @@ def IOU(ann, centroids):
 
     return np.array(similarities)
 
+
 def avg_IOU(anns, centroids):
     n,d = anns.shape
     sum = 0.
@@ -69,6 +69,7 @@ def run_kmeans(ann_dims, anchor_num):
         prev_assignments = assignments.copy()
         old_distances = distances.copy()
 
+
 def generateAnchors(train_annotation_folder, train_image_folder, train_cache_file, model_labels):
 
     print("Generating anchor boxes for training images and annotation...")
@@ -101,7 +102,6 @@ def generateAnchors(train_annotation_folder, train_image_folder, train_cache_fil
     widths = anchors[:, 0]
     sorted_indices = np.argsort(widths)
 
-
     anchor_array = []
     reverse_anchor_array = []
     out_string = ""
@@ -112,7 +112,6 @@ def generateAnchors(train_annotation_folder, train_image_folder, train_cache_fil
 
         out_string += str(int(anchors[i, 0] * 416)) + ',' + str(int(anchors[i, 1] * 416)) + ', '
 
-
     reverse_anchor_array.append(anchor_array[12:18])
     reverse_anchor_array.append(anchor_array[6:12])
     reverse_anchor_array.append(anchor_array[0:6])

imageai/Detection/Custom/voc.py

@@ -1,30 +1,30 @@
-import numpy as np
 import os
 import xml.etree.ElementTree as ET
 import pickle
 
+
 def parse_voc_annotation(ann_dir, img_dir, cache_name, labels=[]):
     if os.path.exists(cache_name):
         with open(cache_name, 'rb') as handle:
             cache = pickle.load(handle)
         all_insts, seen_labels = cache['all_insts'], cache['seen_labels']
     else:
-        all_insts = []
-        seen_labels = {}
+        all_insts = list()
+        seen_labels = dict()
 
         for ann in sorted(os.listdir(ann_dir)):
-            img = {'object':[]}
+            img = {'object': list()}
 
             try:
-                tree = ET.parse(ann_dir + ann)
+                tree = ET.parse(os.path.join(ann_dir, ann))
             except Exception as e:
                 print(e)
-                print('Ignore this bad annotation: ' + ann_dir + ann)
+                print('Ignore this bad annotation: ' + os.path.join(ann_dir, ann))
                 continue
 
             for elem in tree.iter():
                 if 'filename' in elem.tag:
-                    img['filename'] = img_dir + elem.text
+                    img['filename'] = os.path.join(img_dir, elem.text)
                 if 'width' in elem.tag:
                     img['width'] = int(elem.text)
                 if 'height' in elem.tag:
@@ -64,4 +64,4 @@ def parse_voc_annotation(ann_dir, img_dir, cache_name, labels=[]):
         with open(cache_name, 'wb') as handle:
             pickle.dump(cache, handle, protocol=pickle.HIGHEST_PROTOCOL)    
 
-    return all_insts, seen_labels
+    return all_insts, seen_labels