Skip to content
No description, website, or topics provided.
Python
Branch: master
Clone or download

Latest commit

Fetching latest commit…
Cannot retrieve the latest commit at this time.

Files

Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
datasets
demo
README.md Update README.md Dec 6, 2019
datasets.py
fgvc_datasets.py first commit Aug 13, 2019
ft_resnet.py first commit Aug 13, 2019
hooks.py first commit Aug 13, 2019
illustration.png
losses.py first commit Aug 13, 2019
meters.py first commit Aug 13, 2019
optimizers.py first commit Aug 13, 2019
sss_net.py first commit Aug 13, 2019
trainer.py
utility.py first commit Aug 13, 2019
utils.py first commit Aug 13, 2019

README.md

Selective Sparse Sampling for Fine-grained Image Recognition

Illustration

PyTorch Implementation

This repository contains:

  • the pytorch implementation of Selective Sparse Sampling.
  • the CUB-200-2011 demo (training, test).

Please follow the instruction below to install it and run the experiment demo.

Prerequisites

Installation

  1. Install S3N via Nest's CLI tool:

    # note that data will be saved under your current path
    $ git clone https://github.com/Yao-DD/S3N.git ./S3N
    $ nest module install ./S3N/ s3n
    # verify the installation
    $ nest module list --filter s3n

Prepare Data

  1. Download the CUB-200-2011 dataset:

    $ mkdir ./S3N/datasets
    $ cd ./S3N/datasets
    # download and extract data
    $ wget http://www.vision.caltech.edu/visipedia-data/CUB-200-2011/CUB_200_2011.tgz
    $ tar xvf CUB_200_2011.tgz
  2. Prepare annotation files:

    Move the file ./datasets/train.txt and ./datasets/test.txt into ./datasets/CUB_200_2011. The list of image file names and label is contained in the file ./datasets/CUB_200_2011/train.txt and ./datasets/CUB_200_2011/test.txt, with each line corresponding to one image:

    <image_name> <class_id>  
    

Run the demo

  1. run the code as:

    $ cd ./S3N
    # run baseline
    $ PYTHONWARNINGS='ignore' CUDA_VISIBLE_DEVICES=0,1 nest task run ./demo/cub_baseline.yml
    # run S3N
    $ PYTHONWARNINGS='ignore' CUDA_VISIBLE_DEVICES=0,1 nest task run ./demo/cub_s3n.yml

Pretrained models

  1. S3N model for CUB_200_2011 dataset is availavble on Baidu Disk.

    The link:https://pan.baidu.com/s/19x9zI_ZNi32sRGRgNwN_Fw
    code: r252

CAUTION

The current code was prepared under the above-mentioned prerequisites. The use of other version can cause problems.

You can’t perform that action at this time.