DF²Net: A Discriminative Feature Learning and Fusion Network for RGB-D Indoor Scene Classification

This is the code for the paper DF²Net: A Discriminative Feature Learning and Fusion Network for RGB-D Indoor Scene Classification(AAAI 2018).

Installation

1. Install prerequsites for caffe. The caffe version we use can be found in our caffe_dev repo.
2. Add the loss layers to the caffe_dev source and include path.
3. Compile the caffe_dev Github submodule

Our Trained Models

You can download our trained model on SUN RGB-D Dataset and NYU Dataset V2 at Baidu Netdisk or Google Driver.

Our Converted HHA Images

We use the Saurabh Gupta's code to convert the depth images. We also release our converted HHA images of SUN RGB-D Dataset and NYU Depth V2 Dataset at Google Driver.