Datasets:
Languages:
English
ArXiv:
Tags:
egocentric-vision
exocentric-vision
gaze-tracking
referential-expressions
cooking
spatial-reasoning
License:
File size: 5,226 Bytes
537c1f5 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 | """
Look and Tell Referential Dataset Loader
Example script for loading and using the dataset.
"""
from pathlib import Path
import pandas as pd
import json
import cv2
from typing import Dict, Any, Optional
class ARIAReferentialDataset:
"""
Loader for KTH-ARIA Referential Dataset
Example usage:
dataset = ARIAReferentialDataset('data')
recording = dataset.load_recording('par_01', 'rec_01')
"""
def __init__(self, data_path: str):
self.data_path = Path(data_path)
self.metadata = self._load_metadata()
self.recipes = self._load_recipes()
def _load_metadata(self) -> pd.DataFrame:
"""Load dataset metadata"""
metadata_path = self.data_path / 'manifests' / 'metadata.parquet'
if metadata_path.exists():
return pd.read_parquet(metadata_path)
else:
# Fallback to CSV
csv_path = self.data_path / 'manifests' / 'metadata.csv'
return pd.read_csv(csv_path)
def _load_recipes(self) -> Dict:
"""Load recipe information"""
recipe_path = self.data_path / 'manifests' / 'recipes.json'
with open(recipe_path) as f:
return json.load(f)
def get_recording_info(self, participant_id: str, recording_id: str) -> Dict[str, Any]:
"""Get metadata for a specific recording"""
recording_uid = f"{participant_id}_{recording_id}"
row = self.metadata[self.metadata['recording_uid'] == recording_uid]
if len(row) == 0:
raise ValueError(f"Recording {recording_uid} not found")
return row.iloc[0].to_dict()
def load_recording(self, participant_id: str, recording_id: str) -> Dict[str, Any]:
"""
Load all data for a recording
Returns:
Dictionary containing:
- ego_video_path: Path to egocentric video
- exo_video_path: Path to exocentric video
- audio_path: Path to audio
- gaze: Gaze tracking dataframe
- references: Reference annotations dataframe
- transcription: ASR transcription dataframe
- recipe: Recipe information
- metadata: Recording metadata
"""
# Get paths
raw_path = self.data_path / participant_id / 'raw' / recording_id
ann_path = self.data_path / participant_id / 'annotations' / 'v1' / recording_id
# Get metadata
info = self.get_recording_info(participant_id, recording_id)
# Derive recipe_id from recording_num (recording_num 1 -> recipe_01, etc.)
recipe_id = f"recipe_{int(info['recording_num']):02d}"
# Load data
result = {
'ego_video_path': str(raw_path / 'ego_video.mp4'),
'exo_video_path': str(raw_path / 'exo_video.mp4'),
'audio_path': str(raw_path / 'audio.wav'),
'metadata': info,
'recipe': self.recipes.get(recipe_id),
}
# Load gaze data if available
gaze_path = raw_path / 'ego_gaze.csv'
if gaze_path.exists():
result['gaze'] = pd.read_csv(gaze_path)
# Load annotations if available
ref_path = ann_path / 'references.csv'
if ref_path.exists():
result['references'] = pd.read_csv(ref_path)
trans_path = ann_path / 'whisperx_transcription.tsv'
if trans_path.exists():
result['transcription'] = pd.read_csv(trans_path, sep='\t')
return result
def get_recordings_by_recipe(self, recipe_id: str) -> pd.DataFrame:
"""Get all recordings for a specific recipe
Args:
recipe_id: Recipe ID string (e.g., 'recipe_01', 'recipe_02', etc.)
"""
# Extract recipe number from recipe_id (e.g., 'recipe_01' -> 1)
recipe_num = int(recipe_id.split('_')[1])
return self.metadata[self.metadata['recording_num'] == recipe_num]
def get_participant_recordings(self, participant_id: str) -> pd.DataFrame:
"""Get all recordings for a specific participant"""
return self.metadata[self.metadata['participant_id'] == participant_id]
# Example usage
if __name__ == "__main__":
# Initialize dataset
dataset = ARIAReferentialDataset('data')
# Print dataset summary
print(f"Total recordings: {len(dataset.metadata)}")
print(f"Participants: {dataset.metadata['participant_id'].nunique()}")
print(f"Recipes: {len(dataset.recipes) - 1}") # -1 for surface_map
print()
# Load a specific recording
print("Loading par_01, rec_01...")
recording = dataset.load_recording('par_01', 'rec_01')
print(f"Recipe: {recording['recipe']['name']}")
print(f"Duration: {recording['metadata']['duration_sec']:.1f}s")
print(f"Has gaze: {recording['metadata']['has_gaze']}")
print(f"References: {recording['metadata']['n_references']}")
print()
# Get all recordings for recipe 1
recipe_1 = dataset.get_recordings_by_recipe('recipe_01')
print(f"Recipe 1 performed by {len(recipe_1)} participants") |