https://github.com/bes-dev/mpl.pytorch

Pytorch implementation of MaxPoolingLoss.
https://github.com/bes-dev/mpl.pytorch

Last synced: about 2 months ago
JSON representation

Pytorch implementation of MaxPoolingLoss.

• Host: GitHub
• URL: https://github.com/bes-dev/mpl.pytorch
• Owner: bes-dev
• License: bsd-3-clause
• Created: 2018-04-20T22:04:24.000Z (over 6 years ago)
• Default Branch: master
• Last Pushed: 2018-06-09T12:04:20.000Z (over 6 years ago)
• Last Synced: 2024-05-23T06:48:17.986Z (8 months ago)
• Language: Python
• Size: 13.7 KB
• Stars: 175
• Watchers: 6
• Forks: 11
• Open Issues: 4
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

• Awesome-pytorch-list-CNVersion - mpl.pytorch

README

        
# Max-Pooling Loss

[**Loss Max-Pooling for Semantic Image Segmentation**](https://arxiv.org/abs/1704.02966)

## Installation

### Requirements

To install PyTorch, please refer to https://github.com/pytorch/pytorch#installation.

### Compiling

Some parts of Max-Pooling Loss have a native C++ implementation, which must be compiled with the following commands:

```bash

cd mpl

python build.py

```

## Using

```python

import mpl

import torch

max_pooling_loss = mpl.MaxPoolingLoss(ratio=0.3, p=1.7, reduce=True)

loss = torch.Tensor(1, 3, 3, 3).uniform_(0, 1)

loss = max_pooling_loss(loss)

```