ObjectDetector.Dataset.from_voc ZeroDivisionError: float division by zero

Hi,

I try to set up an object detector training with
dataset_train = ObjectDetector.Dataset.from_voc(data_root, splits='trainval')
with data_root structure like this :

Annotations
ImageSets
Main
test.txt
trainval.txt
JPEGImages
classes.txt
And I got the error :
File "Training_v2.py", line 12, in
dataset_train = ObjectDetector.Dataset.from_voc(data_root, splits='trainval')
File "/home/penta/.local/lib/python3.8/site-packages/gluoncv/auto/data/dataset.py", line 486, in from_voc
xmin = max(0, float(xml_box.find('xmin').text) - 1) / width
ZeroDivisionError: float division by zero

My Dataset worked fine with previous version 0.1 of autogluon when I was using
dataset_train = ObjectDetector.Dataset(data_root, classes=('sacs','pots','meubles','cartons','matelas','palette','caddy','bidon','chaise','electromenager',))

I haven't found the detail of the ObjectDetector.Dataset.from_voc function in the documentation so I can't check if this is the correct arguments to give to the function.

Regards.