test_talkingheadbench.py

#!/usr/bin/env python3
"""
Example script for using TalkingHeadBenchFakeTestDataset.

This script demonstrates how to load only fake videos from the test subset
of TalkingHeadBench dataset.
"""

import torch
from torch.utils.data import DataLoader
from src.data.fairtalking_dataset import TalkingHeadBenchFakeTestDataset
from src.data.transforms import build_video_transform

def main():
    # Configuration
    dataset_root = "/apdcephfs_gy4/share_303628665/joywu/dataset/TalkingHeadBench"
    audio_cache_dir = f"{dataset_root}/_audio"
    
    # Video parameters
    num_frames = 16
    frame_stride = 2
    frame_size = 224
    
    # Audio parameters
    audio_seconds = 2.56
    audio_sample_rate = 16000
    
    # DataLoader parameters
    batch_size = 8
    num_workers = 4
    
    # Create dataset
    dataset = TalkingHeadBenchFakeTestDataset(
        root=dataset_root,
        audio_cache_dir=audio_cache_dir,
        generators=["AniPortraitAudio", "EmoPortrait", "Hallo", "Hallo2", 
                   "LivePortrait", "AniPortraitVideo"],
        num_frames=num_frames,
        frame_stride=frame_stride,
        frame_size=frame_size,
        audio_seconds=audio_seconds,
        audio_sample_rate=audio_sample_rate,
        video_transform=build_video_transform(None, training=False),
    )
    
    print(f"Dataset size: {len(dataset)} samples")
    
    # Create DataLoader
    dataloader = DataLoader(
        dataset,
        batch_size=batch_size,
        shuffle=False,
        num_workers=num_workers,
        pin_memory=True,
    )
    
    # Test loading a few batches
    print("Testing dataset...")
    for i, batch in enumerate(dataloader):
        if i >= 3:  # Show first 3 batches
            break
            
        print(f"\nBatch {i+1}:")
        print(f"  Video shape: {batch['video'].shape}")
        print(f"  Audio shape: {batch['audio'].shape}")
        print(f"  Labels: {batch['label'].tolist()}")
        
        # Show generator info for this batch
        generators = [meta['generator'] for meta in batch['meta']]
        print(f"  Generators: {generators}")
        
        # Show basenames
        basenames = [meta['basename'] for meta in batch['meta']]
        print(f"  Basenames: {basenames[:2]}...")  # Show first 2
    
    print(f"\nDataset statistics:")
    print(f"  Total samples: {len(dataset)}")
    
    # Count samples per generator
    generator_counts = {}
    for sample in dataset.samples:
        gen = sample['generator']
        generator_counts[gen] = generator_counts.get(gen, 0) + 1
    
    print(f"  Samples per generator:")
    for gen, count in generator_counts.items():
        print(f"    {gen}: {count}")

if __name__ == "__main__":
    main()