Update albumentations

[glenn-jocher]

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Enhancements in Albumentations data augmentation for YOLOv5.

📊 Key Changes

• 🔍 Added size parameter to Albumentations class constructor and to the classify_albumentations function for dynamic resizing.
• 🖼 Introduced ratio parameter in classify_albumentations to control aspect ratio during random resized crops.
• 🔄 The Albumentations class now passes img_size to the class constructor when called in dataloaders.

🎯 Purpose & Impact

• 📏 The new size parameter allows custom sizing for image augmentation, improving flexibility for different datasets and model requirements.
• 📐 The ratio parameter provides additional control over the aspect ratio of images, which can lead to better performance and more robust models.
• 🌱 These changes can potentially improve YOLOv5's performance by allowing more precise image preprocessing that is better suited to the target use case.