Introduce

UEC FOOD dataset to COCO dataset structure.

Dataset download

UECFOOD101 download

http://foodcam.mobi/dataset100.html

UECFOOD256 download

http://foodcam.mobi/dataset256.html

Split information

For the train-test split we also use:
http://foodcam.mobi/uecfood_split.zip

Split rule

In split information,

Training set = val0.txt, val1.txt, val2.txt
Validation set = val3.txt
Test set = val4.txt

Installation and example usage

git clone https://github.com/Daeil-Jung/UECFOOD_2_COCO
pip install pillow
cd UECFOOD_2_COCO
python uecfood_2_coco.py UECFOOD256 # or UECFOOD100

How to use

python uecfood_2_coco.py [-h] [--path PATH] [--dest DEST] {UECFOOD256,UECFOOD100}

Transform UEC FOOD dataset like COCO dataset structure

positional arguments:

{UECFOOD256,UECFOOD100}: Choose UECFOOD256 or UECFOOD100

optional arguments:

Header	Description
-h, --help	Show this help message and exit
--path PATH, -p PATH	Path of target dataset
--dest DEST, -d DEST	Where you make coco dataset

Default tree structure

├UECFOOD_2_COCO
├─dataset256 # or dataset100
│  └─UECFOOD256
│      ├─1.jpg
│      ├─10.jpg
│      ├─100.jpg
│      ├─ ...
├uecfood_split
│  ├─uecfood100_split
│  └─uecfood256_split
│      ├─val0.txt
│      ├─val1.txt
│      ├─val2.txt
│      ├─val3.txt
│      ├─val4.txt
├─uecfood_2_coco.py

Output

├uecfood256_coco # or uecfood100_coco
├─classes.txt
├─*.jpg
├─annotations
│  ├─train_anno.json
│  ├─test_anno.json
│  └─valid_anno.json