This fork of Caffe supports the various projects on semantic-, instance- and panoptic-segmentation done by the Torr Vision Group.
These include:
-
Conditional Random Fields as Recurrent Neural Networks, ICCV 2015.
-
Higher Order Conditional Random Fields in Deep Neural Networks, ECCV 2016.
-
Bottom-up Instance Segmentation with Deep Higher-Order CRFs, BMVC 2016.
-
Pixelwise Instance Segmentation with a Dynamically Instantiated Network, CVPR 2017.
-
Holistic, Instance-Level Human Parsing, BMVC 2017.
-
On the Robustness of Semantic Segmentation Models to Adversarial Attacks, CVPR 2018, PAMI 2020.
-
Weakly- and Semi-Supervised Panoptic Segmentation. ECCV 2018.
Note that this code base has been refactored and modified since the original papers were written.
Follow the standard guide for installing Caffe.
In addition to the original Caffe, this repository contains layers from Fast-RCNN, R-FCN, Dilated Convolutions, Deeplab, PSPNet, MNC and Intel Caffe.
For any queries, please contact Anurag Arnab ([email protected]). Pull requests are also welcome.