A Pytorch implementation of Pyramid Attention Networks for Semantic Segmentation from 2018 paper by Hanchao Li, Pengfei Xiong, Jie An, Lingxue Wang.

Installation

• Env: Python3.6, Pytorch1.0-preview
• Clone this repository.
• Download the dataset by following the instructions below.

VOC2012 Dataset

The overall dataset is augmented by Semantic Boundaries Dataset, resulting in training data 10582 and test data 1449. https://www.sun11.me/blog/2018/how-to-use-10582-trainaug-images-on-DeeplabV3-code/

After preparing the data, please change the directory below for training.

training_data = Voc2012('/home/tom/DISK/DISK2/jian/PASCAL/VOC2012', 'train_aug', transform=train_transforms)
test_data = Voc2012('/home/tom/DISK/DISK2/jian/PASCAL/VOC2012', 'val',transform=test_transforms)

Evaluation

Pixel acc	mIOU
93.19%	78.498%