Skip to content

An unofficial DeepLabV2 with the pre-train weight of ImageNet.

License

Notifications You must be signed in to change notification settings

johnnylu305/deeplab-imagenet-pytorch

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

32 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

An unofficial Deeplab V2 with the pre-train weight of ImageNet

This repository and codes are largely based on and modified from deeplab-pytorch. I highly recommend visiters to visit the GitHub.

Performance

set CRF mIoU
val O 78.8%
test O 79.1%
set CRF mIoU
val O 77.02%

My Environment

  • Operating System:
    • Ubuntu 16.04.5
  • CUDA:
    • CUDA V10.0.130
  • Nvidia driver:
    • 418.87.01
  • Python:
    • python 3.6.8
  • Python package:
    • tqdm, opencv-python, pydensecrf,...
  • Pytorch:
    • pytorch-gpu 1.4.0

Downloading the VOC12 dataset

Visual Object Classes Challenge 2012 (VOC2012)

Download the VOC12 augmentation dataset

For training

python main.py train --config-path configs/voc12ImageNet.yaml

For testing

python main.py test --config-path configs/voc12Test.yaml

For crf

python main.py crf --config-path configs/voc12Test.yaml --n-jobs 8

OOM

  • Please try small batch size and image size
  • Disable evaluate during training

Further details

kazuto1011/deeplab-pytorch

References

  1. kazuto1011/deeplab-pytorch

  2. L.-C. Chen, G. Papandreou, I. Kokkinos, K. Murphy, A. L. Yuille. DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs. IEEE TPAMI, 2018.
    Project / Code / arXiv paper

  3. H. Caesar, J. Uijlings, V. Ferrari. COCO-Stuff: Thing and Stuff Classes in Context. In CVPR, 2018.
    Project / arXiv paper

  4. M. Everingham, L. Van Gool, C. K. I. Williams, J. Winn, A. Zisserman. The PASCAL Visual Object Classes (VOC) Challenge. IJCV, 2010.
    Project / Paper

About

An unofficial DeepLabV2 with the pre-train weight of ImageNet.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published