https://github.com/minar09/dataset-cfpd-windows
CFPD | Colorful Fashion Parsing Data for Windows OS
https://github.com/minar09/dataset-cfpd-windows
cfpd clothing-parsing colorful-fashion dataset fashion-dataset fashion-parsing windows
Last synced: 10 months ago
JSON representation
CFPD | Colorful Fashion Parsing Data for Windows OS
- Host: GitHub
- URL: https://github.com/minar09/dataset-cfpd-windows
- Owner: minar09
- Created: 2018-12-30T16:01:18.000Z (over 7 years ago)
- Default Branch: master
- Last Pushed: 2020-05-24T15:41:02.000Z (about 6 years ago)
- Last Synced: 2025-07-07T19:21:23.162Z (12 months ago)
- Topics: cfpd, clothing-parsing, colorful-fashion, dataset, fashion-dataset, fashion-parsing, windows
- Language: Jupyter Notebook
- Homepage: https://github.com/hrsma2i/dataset-CFPD
- Size: 449 KB
- Stars: 8
- Watchers: 3
- Forks: 1
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
## Disclaimer
This is a modified repository from [dataset-CFPD](https://github.com/hrsma2i/dataset-CFPD). Please refer to the original repository for more details.
# CFPD | Colorful Fashion Parsing Data
This dataset is used in the paper, [(S. Liu, J. Feng, C. Domokos, H. Xu, J. Huang, Z. Hu, & S. Yan. 2014) CFPD | Fashion parsing with weak color-category labels.](https://sites.google.com/site/fashionparsing/home)
## Details
- 2,682 images
- 600 x 400 (height, width)
- pixel-level annotated (segmentation map)
- 23 categories
- 13 **colors**
- `make_label.py` makes the followings from `fashon_parsing_data.mat`.
- `label/bbox.json`: **bounding box** (for object detection not semantic segmentation).
- `label/categories.tsv`
- `category_id`
- `category`
- `label/main_categories.tsv` is selected from `categories.tsv` for object detection.
## Setup Dataset
```
1. install the requirements
pip install -r requirements.txt
2. download zip file (name) from author's GoogleDrive
download.sh
3. rename "data" file manually to "data.zip" or run the command below:
rename data data.zip
4. unzip data.zip manually or follow instructions from here (http://stahlworks.com/dev/index.php?tool=zipunzip) to unzip from cmd
5. make label/categories.tsv, label/bbox.json from fashon_parsing_data.mat
python make_label.py
6. Convert .mat to hdf5(.h5), run convert_tmm_to_hdf5.m with matlab
7. Convert .hdf5 files into images with annotations and lists
python export.py
```
## fashon_parsing_data.mat
### structure
- #refs#
- 0, 0A~zz: 2,719 groups (each record may be a image info)
- category_label
- np.array, float64(actually int), (1, 425)
- map: super-pix id (1~425) -> fine category id (1~117)
- color_label
- np.array, float64(actually int), (1,425)
- map: super-pix id (1~425) -> fine color id (1~60)
- ?img_name
- np.array, uint16, (2~9 etc, 1)
- range: 59~108 etc
- ???
- segmentation
- np.array, float32, (400, 600)
- map: img pix loc (w, h) -> super-pix id (1~425)
- (height, width) (transposing this) is the right orientation.
- b~x: 23 datasets (each record means category id)
- np.array, uint16, (1, #fine_category)
- range: 1~117
- fine categories's ids under the cateogry.
- y,z,A~K: 13 datasets (each record means color id)
- np.array, uint16, (1, #fine_color)
- range: 1~60
- fine color's ids under the color.
- all_category_name
- np.array, h5py.h5r.Reference, (1, 23)
- Each reference correspods the above cateogory keys under #refs#.
- all_color_name
- np.array, h5py.h5r.Reference, (1, 13)
- Each reference correspods the above color keys under #refs#.
## Dataset Problem
- Mentioned in this paper's Figure 7, [(P. Tangseng, Z. Wu, & K. Yamaguchi. 2017) Looking at Outfit to Parse Clothing.](https://arxiv.org/pdf/1703.01386.pdf).
- [This repository](https://github.com/hrsma2i/fashion-parsing/tree/master/data/tmm_dataset_sharing) is the code of the above paper. This code deals with CFPD.