README.md

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:

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)