Amodal-Instance-Segmentation-through-KINS-Dataset icon indicating copy to clipboard operation
Amodal-Instance-Segmentation-through-KINS-Dataset copied to clipboard

More train ,test and evaluation description?

Open jiangerxiaozhao opened this issue 6 years ago • 2 comments

A very nice job,how can I reproduce Mask R-CNN and Mask R-CNN + ASN on KINS dataset in paper,can you provide related commands and more detailedinstructions? Thank you.

jiangerxiaozhao avatar Oct 11 '19 12:10 jiangerxiaozhao

"ModuleNotFoundError: No module named '_init_paths'", I use https://github.com/roytseng-tw/Detectron.pytorch/blob/master/tools/_init_paths.py but "ModuleNotFoundError: No module named 'pycocotools.coco_amodalinmodalseg'", this means that you have modified the official coco api,but I can't find related code

jiangerxiaozhao avatar Oct 11 '19 13:10 jiangerxiaozhao

Firstly, the reference code is too slow because there was no maskrcnn-benchmark or mmdetection at that time. I recommend extracting some useful modules into the other frameworks.

Secondly, I indeed modified the cocoapi to evaluate amodal and inmodal mask at the same time. It does not matter if you use official cocoapi for training model.

qqlu avatar Oct 14 '19 04:10 qqlu