add vrs download

.gitignore

@@ -5,4 +5,10 @@
 .env.test.local
 .env.production.local
 
-node_modules
+node_modules
+
+# Aria Everyday Activities Dataset files
+Aria Everyday Activities Dataset.json
+
+# Downloaded video files
+aea/

README.md

@@ -28,13 +28,32 @@ task_categories:
 
 This repository contains the SAVVY-Bench, the first benchmark for dynamic 3D spatial reasoning in audio-visual environments, introduced in [SAVVY: Spatial Awareness via Audio-Visual LLMs through Seeing and Hearing](https://arxiv.org/pdf/2506.05414)
 
-
-## Files
+## Video and Audio Files
 
 ```python
 from datasets import load_dataset
 dataset = load_dataset("ZijunCui/SAVVY-Bench")
 ```
+
+## Download Aria Everyday Activities Videos
+
+To use SAVVY-Bench, you need to download the corresponding video files from the Aria Everyday Activities dataset.
+
+### Step 1: Access the Dataset
+1. Visit [Aria Everyday Activities Dataset](https://www.projectaria.com/datasets/aea/)
+2. Follow the instructions to access the dataset
+3. Download the `Aria Everyday Activities Dataset.json` file
+
+### Step 2: Download Required Videos
+
+```bash
+python download_aea_videos.py
+```
+
+### Step 3: Verify Downloads
+After downloading, you should have 53 `.vrs` files in the `aea/vrs/` directory.
+
+
 <!-- 
 ## Dataset Description
 

dataset_infos.json

@@ -0,0 +1,14 @@
+{
+    "dataset_name": "savvy-bench",
+    "description": "SAVVY-Bench dataset for evaluation",
+    "license": "cc-by-nc-sa-4.0",
+    "homepage": "https://github.com/ZijunCui/SAVVY-Bench",
+    "citation": "",
+    "configs": {
+      "default": {
+        "data_files": {
+          "train": "savvy_bench.jsonl"
+        }
+      }
+    }
+  }

download_aea_videos.py

@@ -0,0 +1,100 @@
+#!/usr/bin/env python3
+"""
+Download Aria Everyday Activities videos for SAVVY-Bench
+
+This script downloads the main_vrs files for the video IDs specified in video_ids.txt
+from the Aria Everyday Activities dataset.
+
+Usage:
+    python download_aea_videos.py
+"""
+
+import json
+import os
+import requests
+from pathlib import Path
+
+def download_aea_videos():
+    """Download Aria Everyday Activities videos for SAVVY-Bench"""
+    
+    # Check if required files exist
+    if not os.path.exists('Aria Everyday Activities Dataset.json'):
+        print("Error: 'Aria Everyday Activities Dataset.json' not found.")
+        print("Please download it from https://www.projectaria.com/datasets/aea/")
+        return
+    
+    if not os.path.exists('video_ids.txt'):
+        print("Error: 'video_ids.txt' not found.")
+        return
+    
+    # Load the dataset index
+    print("Loading dataset index...")
+    with open('Aria Everyday Activities Dataset.json', 'r') as f:
+        dataset = json.load(f)
+    
+    # Load video IDs
+    print("Loading video IDs...")
+    with open('video_ids.txt', 'r') as f:
+        video_ids = [line.strip() for line in f.readlines()]
+    
+    print(f"Found {len(video_ids)} video IDs to download")
+    
+    # Create output directory
+    os.makedirs('aea/vrs', exist_ok=True)
+    print("Created directory: aea/vrs")
+    
+    # Download videos
+    downloaded_count = 0
+    skipped_count = 0
+    error_count = 0
+    
+    for i, video_id in enumerate(video_ids, 1):
+        print(f"\n[{i}/{len(video_ids)}] Processing {video_id}...")
+        
+        if video_id in dataset['sequences']:
+            vrs_info = dataset['sequences'][video_id]['main_vrs']
+            filename = vrs_info['filename']
+            download_url = vrs_info['download_url']
+            
+            output_path = f'aea/vrs/{filename}'
+            
+            if os.path.exists(output_path):
+                print(f"  ✓ {filename} already exists (skipped)")
+                skipped_count += 1
+            else:
+                try:
+                    print(f"  ↓ Downloading {filename}...")
+                    response = requests.get(download_url, stream=True)
+                    response.raise_for_status()
+                    
+                    with open(output_path, 'wb') as f:
+                        for chunk in response.iter_content(chunk_size=8192):
+                            f.write(chunk)
+                    
+                    print(f"  ✓ Downloaded {filename}")
+                    downloaded_count += 1
+                    
+                except Exception as e:
+                    print(f"  ✗ Error downloading {filename}: {e}")
+                    error_count += 1
+        else:
+            print(f"  ✗ Video ID {video_id} not found in dataset")
+            error_count += 1
+    
+    # Summary
+    print(f"\n{'='*50}")
+    print("DOWNLOAD SUMMARY")
+    print(f"{'='*50}")
+    print(f"Total video IDs: {len(video_ids)}")
+    print(f"Successfully downloaded: {downloaded_count}")
+    print(f"Skipped (already exists): {skipped_count}")
+    print(f"Errors: {error_count}")
+    print(f"Files saved to: aea/vrs/")
+    
+    if downloaded_count + skipped_count == len(video_ids):
+        print("\n✓ All videos are ready for use!")
+    else:
+        print(f"\n⚠ {error_count} videos failed to download. Please check the errors above.")
+
+if __name__ == "__main__":
+    download_aea_videos() 

video_ids.txt

@@ -0,0 +1,52 @@
+loc1_script2_seq1_rec1
+loc1_script2_seq1_rec2
+loc1_script2_seq3_rec1
+loc1_script2_seq3_rec2
+loc1_script2_seq4_rec1
+loc1_script2_seq4_rec2
+loc1_script2_seq6_rec1
+loc1_script2_seq6_rec2
+loc1_script2_seq7_rec1
+loc1_script2_seq7_rec2
+loc1_script2_seq8_rec1
+loc1_script2_seq8_rec2
+loc2_script2_seq1_rec1
+loc2_script2_seq1_rec2
+loc2_script2_seq2_rec1
+loc2_script2_seq2_rec2
+loc2_script2_seq3_rec1
+loc2_script2_seq3_rec2
+loc2_script2_seq4_rec1
+loc2_script2_seq4_rec2
+loc2_script2_seq5_rec1
+loc2_script2_seq5_rec2
+loc2_script2_seq6_rec1
+loc2_script2_seq6_rec2
+loc2_script2_seq8_rec1
+loc2_script2_seq8_rec2
+loc2_script3_seq1_rec2
+loc2_script3_seq2_rec1
+loc2_script3_seq2_rec2
+loc2_script3_seq3_rec1
+loc2_script3_seq4_rec1
+loc2_script3_seq4_rec2
+loc2_script3_seq5_rec1
+loc2_script3_seq5_rec2
+loc3_script2_seq1_rec1
+loc3_script2_seq1_rec2
+loc3_script2_seq3_rec1
+loc3_script2_seq3_rec2
+loc3_script2_seq4_rec1
+loc3_script2_seq4_rec2
+loc3_script2_seq5_rec1
+loc3_script2_seq5_rec2
+loc3_script2_seq7_rec1
+loc3_script2_seq7_rec2
+loc3_script3_seq1_rec1
+loc3_script3_seq1_rec2
+loc3_script3_seq2_rec1
+loc3_script3_seq2_rec2
+loc3_script3_seq4_rec1
+loc3_script3_seq4_rec2
+loc3_script3_seq5_rec1
+loc3_script3_seq5_rec2

video_ids_final.txt

@@ -0,0 +1,59 @@
+loc1_script2_seq1_rec1
+loc1_script2_seq1_rec2
+loc1_script2_seq3_rec1
+loc1_script2_seq3_rec2
+loc1_script2_seq4_rec1
+loc1_script2_seq4_rec2
+loc1_script2_seq6_rec1
+loc1_script2_seq6_rec2
+loc1_script2_seq7_rec1
+loc1_script2_seq7_rec2
+loc1_script2_seq8_rec1
+loc1_script2_seq8_rec2
+loc2_script2_seq1_rec1
+loc2_script2_seq1_rec2
+loc2_script2_seq2_rec1
+loc2_script2_seq2_rec2
+loc2_script2_seq3_rec1
+loc2_script2_seq3_rec2
+loc2_script2_seq4_rec1
+loc2_script2_seq4_rec2
+loc2_script2_seq5_rec1
+loc2_script2_seq5_rec2
+loc2_script2_seq6_rec1
+loc2_script2_seq6_rec2
+loc2_script2_seq8_rec1
+loc2_script2_seq8_rec2
+loc2_script3_seq1_rec2
+loc2_script3_seq2_rec1
+loc2_script3_seq2_rec2
+loc2_script3_seq31_rec1
+loc2_script3_seq31_rec2
+loc2_script3_seq32_rec1
+loc2_script3_seq32_rec2
+loc2_script3_seq33_rec1
+loc2_script3_seq33_rec2
+loc2_script3_seq34_rec1
+loc2_script3_seq34_rec2
+loc2_script3_seq4_rec1
+loc2_script3_seq4_rec2
+loc2_script3_seq5_rec1
+loc2_script3_seq5_rec2
+loc3_script2_seq1_rec1
+loc3_script2_seq1_rec2
+loc3_script2_seq3_rec1
+loc3_script2_seq3_rec2
+loc3_script2_seq4_rec1
+loc3_script2_seq4_rec2
+loc3_script2_seq5_rec1
+loc3_script2_seq5_rec2
+loc3_script2_seq7_rec1
+loc3_script2_seq7_rec2
+loc3_script3_seq1_rec1
+loc3_script3_seq1_rec2
+loc3_script3_seq2_rec1
+loc3_script3_seq2_rec2
+loc3_script3_seq4_rec1
+loc3_script3_seq4_rec2
+loc3_script3_seq5_rec1
+loc3_script3_seq5_rec2