|
--- |
|
dataset_info: |
|
features: |
|
- name: image |
|
dtype: image |
|
- name: conditioning_image |
|
dtype: image |
|
- name: text |
|
dtype: string |
|
splits: |
|
- name: train |
|
num_bytes: 111989279184.95 |
|
num_examples: 507050 |
|
download_size: 112032639870 |
|
dataset_size: 111989279184.95 |
|
--- |
|
# Dataset Card for "hagrid-mediapipe-hands" |
|
|
|
This dataset is designed to train a ControlNet with human hands. It includes hand landmarks detected by MediaPipe(for more information refer to: https://developers.google.com/mediapipe/solutions/vision/hand_landmarker). |
|
The source image data is from [HaGRID dataset](https://github.com/hukenovs/hagrid) and we use a modified version from Kaggle(https://www.kaggle.com/datasets/innominate817/hagrid-classification-512p) to build this dataset. There are 507050 data samples in total and the image resolution is 512x512. |
