Fine Grained Image Classification on CUB-200-2011
We suggest using Anaconda to create a virtual environment for this program. Visit official website or here to download the installer(Hope there is a GUI and a browser on your deep learning server machine).
Create a new virtual environment:
conda create -n pytorch python=3.6
Activate the environment on MacOS/Linux:
source activate pytorch
On Windows:
activate pytorch
Note: We suggest using pip
instead of conda
to install following requirements on Windows. The reason is that if you choose to use conda to install something like PyTorch or numpy, in order to speed up computation, another 3 packages start with mkl
will also be downloaded. However, these mkl
packages have conflicts with conda
on Windows and you just cannot run the program.
If you're using MacOS or Linux, just ignore the note and enjoy conda
~
If you want to speed up package download, you can add Tsinghua's package repository for conda
:
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free/
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main/
conda config --set show_channel_urls yes
Visit 清华大学开源软件镜像站 for more information.
Visit Official Website, choose correct OS/PM/Python-version/CUDA-version to get install command. Please install both pytorch
and torchvision
.
If your download speed is too slow, you can also add Tsinghua's repository specially for installing pytorch:
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/pytorch/
conda install pytorch torchvision
pip install requests
, note that conda
doesn't contain this package.
conda install matplotlib pillow
cd FGC_CUB-200-2011
source activate pytorch
python global.py