add header=None to pd.read_csv #148

Hi just wanna report a small mistake in YOLO dataset.py

 class VOCDataset(...):
    def __init__(...):
        self.annotations = pd.read_csv(csv_file)

should be self.annotations = pd.read_csv(csv_file, header=None), otherwise the first line of the file is used as header and thus not used in training/testing