This code is a re-implementation of the following paper: Mengran Gou, Fei Xiong, Octavia Camps, Mario Sznaier. "MoNet: Moment embedding network", CVPR 2018
The code is tested on Ubuntu 14.04 using a single NVIDIA Titan X GPU, 64G RAM and MATLAB R2016b.
- VGG-VG16 model
How to run
- Download the pretrained VGG model and put it under ./data/models
- Download one dataset from the following list and put it under ./data
- Modify the paths in './base/setup.m'
- Run "script_run_exp.m". The default setting will train/test MoNet on CUB
- If you kill the code during the pre-trained feature extraction stage, you have to delet ./data/%NETWORK%/%DATASET%/nonftGlbGau and then re-start
This code is modified from the G2DeNet package and based on BCNN code. Please refer to the originial websites for more details