|Project Name||Stars||Downloads||Repos Using This||Packages Using This||Most Recent Commit||Total Releases||Latest Release||Open Issues||License||Language|
|Panopticsegmentation||71||4 years ago||2||Python|
|PKU_360 Team's fuse code! This code is used to fuse the semantic segmentation result and instance segmentation result. We won the third place in COCO2018 panoptic segmentation.|
|Multimodal_segmentation||5||3 years ago||3||mit||Python|
|Code for Disentangle Align and Fuse for Multimodal and Zero-shot Image Segmentation|
|A Cnn Model For Human Parsing Based On Capacity Optimization||1||4 years ago||Python|
This repo provides the code to combine the semantic and instance segmentation predictions for panoptic segmentation two-channel-pngs output. Compared with the official version, it tries to solve occlusion problems and takes some object relationships into account, according to the method of the third place winner in the COCO2018 panoptic segmentation competition. A detailed description can be found here.
I advise you to organize you file in the following ways.
PanopticSegmentation ├─files ├─annotations |─detresults
You can form the file organization by the tools generate_file_organization.py. It will generate the file structure automatically.
cd REPO_ROOT_DIR python tools/generate_file_organization.py
python tools/cal_panoptic.py -t
The evaluation code is from the official panopticapi
If you want to evaluate the result after generating two-channel-pngs automatically, just need to add '-e'. If the two-channel-pngs have been generated and you only want to evaluate, then just add '-sc'.
python tools/cal_panoptic.py -e # generate two-channel-pngs and automatically evaluate the result.
python tools/cal_panoptic.py -sc # only run evaluation.
If you have any question, just leave an issue!