Single class train update

[glenn-jocher]

Fix for 'item' class appearing on forced --single-cls training of existing single class datasets. See #12 (comment)

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Refinement of single-class mode training in YOLOv5.

📊 Key Changes

• Tweaked the logic setting up the number of classes (nc) and the names of the classes (names) for single-class datasets in train.py.
• Amended the help description for the --single-cls argument to clarify its purpose.

🎯 Purpose & Impact

• 🎯 The adjustment ensures that single-class training mode correctly handles the class names within the dataset, whether there is only one name or multiple names present.
• 🚀 This update improves usability by providing clear guidance when training YOLOv5 on single-class datasets, potentially preventing errors and confusion for users.
• 📈 Users opting to train models with a single class can expect more robust behavior, especially when the dataset might inaccurately have multiple names listed.