Deep Individual Active Learning: Safeguarding against Out-of-Distribution Challenges in Neural Networks