- Make a clone of this repo.
git clone https://github.com/JaehaKim97/SR4IR.git
cd SR4IR
- Set the environment. We recommend using Anaconda.
conda env create -f assets/environment.yaml
conda activate SR4IR
python src/setup.py
If you encounter an error, please try another installation command as noted here.
We use the Pascal-VOC dataset. You can download it from the official homepage.
Locate the dataset file VOCtrainval_11-May-2012.tar
at datasets/
then run the below command:
python preprocess/voc/main.py
The processed dataset will located at datasets/VOC
.
We used the Pascal-VOC dataset again, but with some pre-processing steps that convert the dataset's annotation into the COCO style.
Make sure that you follow the above dataset preparation for semantic segmentation, i.e., you have the VOC dataset within the path: datasets/VOC
.
Then execute the below command:
bash preprocess/voc2coco/main.sh
It will generate COCO style annotated VOC dataset within the path datasets/VOC_COCO
We use two datasets; StanfordCars and CUB-200-2011 datasets.
Download the dataset via Kaggle in here (Download button in top-right).
Locate the dataset file archive.zip
at datasets/
then run the below command:
python preprocess/stanfordcars/main.py
The processed dataset will located at datasets/StanfordCars
.
Download the dataset via Kaggle in here (Download button in top-right).
Locate the dataset file archive.zip
at datasets/
then run the below command:
python preprocess/cub-200-2011/main.py
The processed dataset will located at datasets/CUB200
.