SignX/smkd/check_data.py

import numpy as np
import os
from pathlib import Path

# Load training data
train_info = np.load('asllrp/train_info.npy', allow_pickle=True).item()

print("=== TRAINING DATA STRUCTURE ===")
print(f"Number of training samples: {len(train_info) - 1}")  # -1 for 'prefix' key
print(f"Prefix: {train_info.get('prefix', 'N/A')}")

# Check a few samples
for i in range(min(5, len(train_info) - 1)):
    sample = train_info[i]
    print(f"\nSample {i}:")
    for key in sample.keys():
        val = sample[key]
        if key == 'label':
            print(f"  {key}: {val} (length: {len(val) if hasattr(val, '__len__') else 'N/A'})")
        elif key == 'folder':
            # Check if the folder exists
            folder_path = Path(val)
            exists = folder_path.exists()
            if exists:
                jpg_files = list(folder_path.glob('*.jpg'))
                print(f"  {key}: {val} (exists: {exists}, jpg files: {len(jpg_files)})")
            else:
                print(f"  {key}: {val} (exists: {exists})")
        else:
            print(f"  {key}: {val}")

# Check if video folders exist
print("\n=== CHECKING VIDEO FOLDER PATHS ===")
missing_folders = []
for i in range(min(10, len(train_info) - 1)):
    sample = train_info[i]
    folder = sample.get('folder', '')
    if not Path(folder).exists():
        missing_folders.append(folder)

if missing_folders:
    print(f"WARNING: {len(missing_folders)} out of 10 sample folders are missing!")
    print("First missing folder:", missing_folders[0] if missing_folders else "None")
else:
    print("All checked folders exist")