TorchSeg：基于pytorch的语义分割算法开源了

加入极市专业CV交流群，与6000+来自腾讯，华为，百度，北大，清华，中科院等名企名校视觉开发者互动交流！更有机会与李开复老师等大牛群内互动！

同时提供每月大咖直播分享、真实项目需求对接、干货资讯汇总，行业技术交流。点击文末“阅读原文”立刻申请入群~

作者：ycszen

原文：https://zhuanlan.zhihu.com/p/55563637

项目地址：https://github.com/ycszen/TorchSeg#why-this-name-furnace

Hi, All

我们开源了TorchSeg，其中包括我们复现的算法：DFN, BiSeNet, PSPNet.

TorchSeg

This project aims at providing a fast, modular reference implementation for semantic segmentation models using PyTorch.

Highlights

• Modular Design: easily construct a customized semantic segmentation models by combining different components.

• Distributed Training: >60% faster than the multi-thread parallel method(nn.DataParallel), we use the multi-processing parallel method.

• Multi-GPU training and inference: support different manners of inference.

• Provides pre-trained models and implement different semantic segmentation models.

Prerequisites

• PyTorch 1.0

• pip3 install torch torchvision

• Easydict

• pip3 install easydict

• Apex

• Ninja

• sudo apt-get install ninja-build

• tqdm

• pip3 install tqdm

Model Zoo

Supported Model


• FCN

• DFN

• BiSeNet

• PSPNet

Performance and Benchmarks

SS:Single Scale MSF:Multi-scale + Flip

PASCAL VOC 2012

80.61: this result reported in paper is further finetuned on train dataset.

Cityscapes

Non-real-time Methods

BiSeNet(ours)1: because we didn't pre-train the Xception39 model on ImageNet in PyTorch, we train this experiment from scratch. We will release the pre-trained Xception39 model in PyTorch and the corresponding experiment.

Real-time Methods

ADE

后续我们会开源更多复现的语义分割算法，欢迎大家 Star 和 Contribute~

*延伸阅读

---

每月大咖直播分享、真实项目需求对接、干货资讯汇总，行业技术交流。点击左下角“阅读原文”立刻申请入群~