metadata
license: cc
dataset_info:
features:
- name: mask
dtype: image
- name: segments_info
struct:
- name: file_name
dtype: string
- name: image_id
dtype: int64
- name: segments_info
list:
- name: area
dtype: float64
- name: category_id
dtype: int64
- name: id
dtype: int64
- name: iscrowd
dtype: int64
- name: isthing
dtype: int64
- name: image_info
struct:
- name: coco_url
dtype: string
- name: date_captured
dtype: string
- name: file_name
dtype: string
- name: height
dtype: int64
- name: id
dtype: int64
- name: license
dtype: int64
- name: width
dtype: int64
splits:
- name: train
num_bytes: 1799961313.342
num_examples: 241602
download_size: 1641153199
dataset_size: 1799961313.342
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
How to download
- Set up environment
pip install datasets tqdm
wget https://raw.githubusercontent.com/bytedance/coconut_cvpr2024/main/download_coconut.py
- Use the download script to download the COCONut dataset splits.
python download_coconut.py --split coconut_b # default split: relabeled_coco_val, need to switch to coconut_b
- Download other COCONut dataset splits.
If you want to download the other splits, you can replace the split name to "relabeled_coco_val" or "coconut_s" NOTE: multiple splits download is not yet supported.
python download_coconut.py --split relabeled_coco_val --output_dir relabeled_coco_val
- The mask images are nearly black as we use continuous segment ids for each image, you can use github visualization tutorial to create colorful masks for viewing.
Please go to our offical github repo for detailed usage instruction: https://github.com/bytedance/coconut_cvpr2024