| 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) | |