This project preprocess the Clothing 1M Dataset and gives a classwise directory structure. The raw dataset can be obtained by following steps on official repository
Check using find test/ -maxdepth 2 -type f | wc -l
in terminal
- Number of nosiy train images - 1000000
- Number of clean train images - 47570
- Number of val images - 10526
- Number of test images - 14313
The images in the directroy for each partition of the dataset are arranged in the directory with q00xx where xx is the class number from 0-13.
└── Clothing1M
├── clean_train
| ├── q0000
| ├── q0001
| └── ...
| └── q0013
|
├── noisy_train
| ├── q0000
| ├── q0001
| └── ...
| └── q0013
|
├── val
| ├── q0000
| ├── q0001
| └── ...
| └── q0013
|
├── test
| ├── q0000
| ├── q0001
| └── ...
| └── q0013
|
├── xxxx
├── xxxx
├── xxxx
├── xxxx
├── xxxx
└── creat_dataset.sh
└── helper.py
└── Readme.md
└── Citation.cff
└── LICENSE
-
Get access to dataset following steps on official repository
-
Clone this repository.
git clone https://github.com/sangamesh-kodge/Clothing1M.git
-
Download the dataset from the download link obtained in step 1 in the cloned repository from step 2. (Or move data after download. )
-
Unzip files in
images/
directory usingtar -xf <file_name.tar>
. Additionally unzip file in the root directory usingunzip <file_name.zip>
. (See the commented lines 20-29 in create_dataset.sh). At this stage the cloned repository should haveimages/
directory containing raw images and files with the label information for each image. -
Run the following command in your terminal/command contraining the cloned repository
sh create_dataset.sh
The project preprocess Clothing1M Dataset and gives a classwise directory structure.
Kindly cite the repository if you use the code. Thanks!
Kodge, S. (2024). Clothing1M [Computer software]. https://github.com/sangamesh-kodge/Clothing1M
@software{Kodge_Clothing1M_2024,
author = {Kodge, Sangamesh},
month = feb,
title = {{Clothing1M}},
url = {https://github.com/sangamesh-kodge/Clothing1M},
year = {2024}
}