https://github.com/pathak22/caffe-ccnn
Branch of caffe for weakly supervised learning.
https://github.com/pathak22/caffe-ccnn
Last synced: about 1 year ago
JSON representation
Branch of caffe for weakly supervised learning.
- Host: GitHub
- URL: https://github.com/pathak22/caffe-ccnn
- Owner: pathak22
- License: other
- Created: 2015-12-09T12:46:18.000Z (over 10 years ago)
- Default Branch: master
- Last Pushed: 2015-12-09T13:28:49.000Z (over 10 years ago)
- Last Synced: 2025-03-24T11:13:33.101Z (over 1 year ago)
- Language: C++
- Size: 27.8 MB
- Stars: 9
- Watchers: 2
- Forks: 2
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# caffe-ccnn
=============
This is the branch of caffe created for [CCNN](https://github.com/pathak22/ccnn) project.
This is a pre-release Caffe branch.
This includes several unmerged PRs and has no guarantees.
Everything here is subject to change, including the history of this branch.
See `future.sh` for details.
Some Key Functionalities:
- Test Time DenseCRF Layer originally described in this [paper](http://graphics.stanford.edu/projects/densecrf/).
- Atrous or 'hole' algorithm described in the [arxiv paper](http://arxiv.org/abs/1412.7062). **Disclaimer**: The Atrous code in this repository is obtained (and modified) from the original implementation, released [here](https://bitbucket.org/deeplab/deeplab-public/).
- A rough version of MIL based Softmax Loss Layer for weakly supervised tasks.
- Background Weighted SoftmaxLossLayer.
- Some additional scripts in the folder ./my_script/