Detectron2 (Original+ResNeSt+centermask2)

This is a modification of the offiical detectron2 colab that can be found here.

This colab is a combination of facebookresearch/detectron2, zhanghang1989/detectron2-ResNeSt and youngwanLEE/centermask2.

Important information

• If you can't open Colab-Detectron2.ipynb inside your Google Drive, try this colab link and save it to your Google Drive. The "open in Colab"-button can be missing in Google Drive, if that person never used Colab.
• Google Colab does assign a random GPU. It depends on luck.
• The Google Colab VM does have a maximum session length of 12 hours. Additionally there is a 30 minute timeout if you leave colab. The VM will be deleted after these timeouts.

Object Detection

Method	Backbone	mAP%	download
Faster R-CNN	ResNet-50	39.25	config | model | log
	ResNet-101	41.37	config | model | log
	ResNeSt-50 (ours)	42.33	config | model | log
	ResNeSt-50-DCNv2 (ours)	44.11	config | model | log
	ResNeSt-101 (ours)	44.72	config | model | log
Cascade R-CNN	ResNet-50	42.52	config | model | log
	ResNet-101	44.03	config | model | log
	ResNeSt-50 (ours)	45.41	config | model | log
	ResNeSt-101 (ours)	47.50	config | model | log
	ResNeSt-200 (ours)	49.03	config | model | log

We train all models with FPN, SyncBN and image scale augmentation (short size of a image is pickedrandomly from 640 to 800). 1x learning rate schedule is used. All of them are reported on COCO-2017 validation dataset.

Instance Segmentation

Method	Backbone	bbox	mask	download
Mask R-CNN	ResNet-50	39.97	36.05	config | model | log
	ResNet-101	41.78	37.51	config | model | log
	ResNeSt-50 (ours)	42.81	38.14	config | model | log
	ResNeSt-101 (ours)	45.75	40.65	config | model | log
Cascade R-CNN	ResNet-50	43.06	37.19	config | model | log
	ResNet-101	44.79	38.52	config | model | log
	ResNeSt-50 (ours)	46.19	39.55	config | model | log
	ResNeSt-101 (ours)	48.30	41.56	config | model | log
	ResNeSt-200-tricks-3x (ours)	50.54	44.21	config | model | log
	ResNeSt-200-dcn-tricks-3x (ours)	50.91	44.50	config | model | log
		53.30*	47.10*

All models are trained along with FPN and SyncBN. For data augmentation,input images’ shorter side are randomly scaled to one of (640, 672, 704, 736, 768, 800). 1x learning rate schedule is used, if not otherwise specified. All of them are reported on COCO-2017 validation dataset. The values with * demonstrate the mutli-scale testing performance on the test-dev2019.

Panoptic Segmentation

Backbone	bbox	mask	PQ	download
ResNeSt-200	51.00	43.68	47.90	config | model | log

Training and Inference

Please follow INSTALL.md to install detecron2.

To train a model with 8 gpus, please run

python tools/train_net.py  --num-gpus 8 --config-file your_config.yaml

For inference

python tools/train_net.py  \
                --config-file your_config.yaml
                --eval-only MODEL.WEIGHTS /path/to/checkpoint_file

For the inference demo, please see GETTING_STARTED.md.

Reference

ResNeSt: Split-Attention Networks [arXiv]

Hang Zhang, Chongruo Wu, Zhongyue Zhang, Yi Zhu, Zhi Zhang, Haibin Lin, Yue Sun, Tong He, Jonas Muller, R. Manmatha, Mu Li and Alex Smola

@article{zhang2020resnest,
title={ResNeSt: Split-Attention Networks},
author={Zhang, Hang and Wu, Chongruo and Zhang, Zhongyue and Zhu, Yi and Zhang, Zhi and Lin, Haibin and Sun, Yue and He, Tong and Muller, Jonas and Manmatha, R. and Li, Mu and Smola, Alexander},
journal={arXiv preprint arXiv:2004.08955},
year={2020}
}

Related Repo

• ResNeSt (main)
• ResNeSt on ImageNet (Gluon, PyTorch)
• ResNeSt on Semantic Segmentation (Gluon, PyTorch)

Contributors

Chongruo Wu, Zhongyue Zhang, Hang Zhang