search

facebookresearch / maskrcnn-benchmark

Fast, modular reference implementation of Instance Segmentation and Object Detection algorithms in PyTorch.
MIT License
9.3k stars 2.49k forks source link

How to train Cityscapes dataset #1103

Open InstantWindy opened 5 years ago

InstantWindy commented 5 years ago

Hi！How to load and train cityscapes dataset? Cityscapes dataset has no bounding box, so how do you train it for detection? thanks.

ingeechart commented 5 years ago

you can get a bounding box when it changes to coco detection format. first change cityscape format to coco panoptic format using converting script in cityscapes dataset github, and then change coco panoptic dataset to coco detection format using coco/panopticapi/converter. then you can use cocoapi to load data.

botcs commented 5 years ago

Please check out PR #1090 with the update you can load files directly in their original format.

By convention it uses the tightest bounding box for each instance, based on their binary mask.

InstantWindy commented 5 years ago

thanks@ingeechart @botcs

InstantWindy commented 5 years ago

Please check out PR #1090 with the update you can load files directly in their original format.

By convention it uses the tightest bounding box for each instance, based on their binary mask.

Hi. When I run the convert_cityscapes_to_coco.py, it shows error as follows: ModuleNotFoundError: No module named 'cityscapesscripts.evaluation.instances2dict_with_polygons' Do you know how to solve it? Thanks.