What does fine-tuning mean in the context of computer vision models? How is fine-tuning used to improve the performance of pre-trained deep learning models? What are the steps involved in fine-tuning a computer vision model for a specific task? How does fine-tuning reduce training time and data requirements? What are the advantages and challenges of fine-tuning models in real-world computer vision applications?