bugs fixed in train.py and update README

README.md

@@ -30,6 +30,10 @@ python yolo.py   OR   python yolo_video.py
     Box format: x_min,y_min,x_max,y_max,class_id (no space).  
     For VOC dataset, try `python voc_annotation.py`
 
-2. Modify train.py and start training.  
+2. Make sure you have run `python convert.py yolov3.cfg yolov3.weights model_data/yolo.h5`  
+    A file model_data/yolo_weights.h5 will be generated when you run train.py for the first time.  
+    The file is used to load pretrained weights.
+
+3. Modify train.py and start training.  
     `python train.py`  
     You will get the trained model model_data/my_yolo.h5.

train.py

@@ -112,8 +112,8 @@ def create_model(input_shape, anchors, num_classes, load_pretrained=True, freeze
         if not os.path.exists(weights_path):
             print("CREATING WEIGHTS FILE" + weights_path)
             yolo_path = os.path.join('model_data', 'yolo.h5')
-            model_body = load_model(yolo_path, compile=False)
-            model_body.save_weights(weights_path)
+            orig_model = load_model(yolo_path, compile=False)
+            orig_model.save_weights(weights_path)
         model_body.load_weights(weights_path, by_name=True, skip_mismatch=True)
         if freeze_body:
             # Do not freeze 3 output layers.

yolo3/utils.py

@@ -19,8 +19,8 @@ def letterbox_image(image, size):
     '''resize image with unchanged aspect ratio using padding'''
     image_w, image_h = image.size
     w, h = size
-    new_w = int(image_w * min(w/image_w, h/image_h))
-    new_h = int(image_h * min(w/image_w, h/image_h))
+    new_w = int(image_w * min(w*1.0/image_w, h*1.0/image_h))
+    new_h = int(image_h * min(w*1.0/image_w, h*1.0/image_h))
     resized_image = image.resize((new_w,new_h), Image.BICUBIC)
 
     boxed_image = Image.new('RGB', size, (128,128,128))