Figure_Skating_Data / README.md
meghna2801's picture
Upload README.md
4f6601b verified
MMFS Figure Skating Dataset (Processed)
This repository contains processed skeleton data for figure skating action classification.
Files included:
- `skeleton/train_data.pkl` — list of numpy arrays; each sample shape (150, 17, 3, 1)
- `skeleton/test_data.pkl` — same as above for test
- `skeleton/train_label.pkl`, `skeleton/test_label.pkl` — integer class labels
- `skeleton/train_name.pkl`, `skeleton/test_name.pkl` — original action names (strings)
- `skeleton/train_score.pkl`, `skeleton/test_score.pkl` — action scores (floats)
- `train_metadata_v2.csv`, `test_metadata_v2.csv` — metadata with `data_index` and cleaned `action_clean`
- `metadata_clean.csv` — compact metadata for training with `data_path` (if present)
How to load a sample:
```python
import pickle
train_data = pickle.load(open('skeleton/train_data.pkl','rb'))
train_labels = pickle.load(open('skeleton/train_label.pkl','rb'))
i = 0
x = train_data[i] # numpy array, shape (150,17,3,1)
y = train_labels[i]
```
Contact: meghna (meghna2801)