DeepLab2 is a TensorFlow library for deep labeling, aiming to provide a unified and state-of-the-art TensorFlow codebase for dense pixel labeling tasks, including, but not limited to semantic segmentation, instance segmentation, panoptic segmentation, depth estimation, or even video panoptic segmentation.
Deep labeling refers to solving computer vision problems by assigning a predicted value for each pixel in an image with a deep neural network. As long as the problem of interest could be formulated in this way, DeepLab2 should serve the purpose. Additionally, this codebase includes our recent and state-of-the-art research models on deep labeling. We hope you will find it useful for your projects.
-
10/18/2022: Add kMaX-DeepLab ADE20K panoptic segmentation results in model zoo.
-
10/04/2022: Open-source MOAT model code and ImageNet pretrained weights. We thank Chenglin Yang for their valuable contributions.
-
08/26/2022: Add ViP-DeepLab support for Waymo Open Dataset: Panoramic Video Panoptic Segmentation. We thank Jieru Mei, Alex Zhu, Xinchen Yan, and Hang Yan, for their valuable contributions.
-
08/16/2022: Support Colab demo for kMaX-DeepLab.
-
07/12/2022: Open-source k-means Mask Transformer (kMaX-DeepLab) code and model zoo.
-
07/11/2022: Drop support of Tensorflow 2.5. Please update to 2.6.
-
04/27/2022: Add ViP-DeepLab demo and update ViP-DeepLab model zoo.
-
09/07/2021: Add numpy implementation of Segmentation and Tracking Quality. Find it here.
-
09/06/2021: Update Panoptic-DeepLab w/ MobileNetv3 backbone results on Cityscapes.
-
08/13/2021: Open-source MaX-DeepLab-L COCO checkpoints (51.3% PQ on COCO val set).
-
07/26/2021: Add ViP-DeepLab support for SemKITTI-DVPS.
-
07/07/2021: KITTI-STEP and MOTChallenge-STEP are ready to use.
-
06/07/2021: Add hungarian matching support on TPU for MaX-DeepLab, thanks to the help from Jiquan Ngiam and Amil Merchant.
-
06/01/2021: "Hello, World!", DeepLab2 made publicly available.
See Installation.
The dataset needs to be converted to TFRecord. We provide some examples below.
Some guidances about how to convert your own dataset.
We list a few projects that use DeepLab2.
-
kMaX-DeepLab Colab notebook for off-the-shelf inference with COCO checkpoints.
-
Panoptic-DeepLab Colab notebook for off-the-shelf inference with Cityscapes checkpoints.
-
ViP-DeepLab Colab notebook for off-the-shelf inference with Cityscapes-DVPS checkpoints.
See Getting Started. In short, run the following command:
To run DeepLab2 on GPUs, the following command should be used:
python trainer/train.py \
--config_file=${CONFIG_FILE} \
--mode={train | eval | train_and_eval | continuous_eval} \
--model_dir=${BASE_MODEL_DIRECTORY} \
--num_gpus=${NUM_GPUS}
Please check FAQ if you have some questions before
reporting the issues.
- Mark Weber, github: markweberdev
- Huiyu Wang, github: csrhddlam
- Siyuan Qiao, github: joe-siyuan-qiao
- Jun Xie, github: clairexie
- Maxwell D. Collins, github: mcollinswisc
- YuKun Zhu, github: yknzhu
- Liangzhe Yuan, github: yuanliangzhe
- Dahun Kim, github: mcahny
- Qihang Yu, github: yucornetto
- Liang-Chieh Chen, github: aquariusjay
-
Note that this library contains our re-implemented DeepLab models in TensorFlow2, and thus may have some minor differences from the published papers (e.g., learning rate).
-
This is not an official Google product.
If you find DeepLab2 useful for your project, please consider citing
DeepLab2
along with the relevant DeepLab series.
- DeepLab2:
@article{deeplab2_2021,
author={Mark Weber and Huiyu Wang and Siyuan Qiao and Jun Xie and Maxwell D. Collins and Yukun Zhu and Liangzhe Yuan and Dahun Kim and Qihang Yu and Daniel Cremers and Laura Leal-Taixe and Alan L. Yuille and Florian Schroff and Hartwig Adam and Liang-Chieh Chen},
title={{DeepLab2: A TensorFlow Library for Deep Labeling}},
journal={arXiv: 2106.09748},
year={2021}
}
-
Marius Cordts, Mohamed Omran, Sebastian Ramos, Timo Rehfeld, Markus Enzweiler, Rodrigo Benenson, Uwe Franke, Stefan Roth, and Bernt Schiele. "The cityscapes dataset for semantic urban scene understanding." In CVPR, 2016.
-
Andreas Geiger, Philip Lenz, and Raquel Urtasun. "Are we ready for autonomous driving? the kitti vision benchmark suite." In CVPR, 2012.
-
Jens Behley, Martin Garbade, Andres Milioto, Jan Quenzel, Sven Behnke, Cyrill Stachniss, and Jurgen Gall. "Semantickitti: A dataset for semantic scene understanding of lidar sequences." In ICCV, 2019.
-
Alexander Kirillov, Kaiming He, Ross Girshick, Carsten Rother, and Piotr Dollar. "Panoptic segmentation." In CVPR, 2019.
-
Dahun Kim, Sanghyun Woo, Joon-Young Lee, and In So Kweon. "Video panoptic segmentation." In CVPR, 2020.
-
Tsung-Yi Lin, Michael Maire, Serge Belongie, James Hays, Pietro Perona, Deva Ramanan, Piotr Dollar, and C Lawrence Zitnick. "Microsoft COCO: Common objects in context." In ECCV, 2014.
-
Patrick Dendorfer, Aljosa Osep, Anton Milan, Konrad Schindler, Daniel Cremers, Ian Reid, Stefan Roth, and Laura Leal-Taixe. "MOTChallenge: A Benchmark for Single-camera Multiple Target Tracking." IJCV, 2020.
-
Bolei Zhou, Hang Zhao, Xavier Puig, Sanja Fidler, Adela Barriuso, and Antonio Torralba. "Scene Parsing through ADE20K Dataset." In CVPR, 2017.