Skip to content

mboboGO/MOMN

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

16 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

MOMN

This is an simple implementation of Multi-Objective Matrix Normalization for Fine-grained Visual Recognition by Pytorch. Paper is accepted by TIP. The code will be re-organized to make it more clear.

Requirements

pytorch 1.0

Training

The training scripts for CUB, Cars, Air, and Dogs are given in https://drive.google.com/drive/folders/1mgKoXwDg3oUGiJluCSWlZJkvrhsbq2tw?usp=sharing. Other extensions can be easily modified.

A detailed illustration is as follows:

step 1:

adding your data path around the 130 line in main.py

step 2:

creating a running bash scrip, as the given example in the google drive. specifically, the running command should be given by:

python main.py -a momn -d cub -s ./cub/checkpoints --backbone densenet201 -b 230 --lr 0.1 --resize_size 560 --crop_size 512 --epochs 90 --is_fix --pretrained

Testing

An example script for testing is also given in the google drive.

pretrained models

Several pretrained models are given in https://drive.google.com/drive/folders/1mgKoXwDg3oUGiJluCSWlZJkvrhsbq2tw?usp=sharing, of which the performance is listed in the paper.

Releases

No releases published

Packages

No packages published