The spontaneous and reflective reaction of the facial expression due to pain is utilized for the analysis of the pain intensity estimation for the patients like children, adults or infants who are affected by diseases like consciousness disorders, dementia, or neural impairments. The human face is considered as a significant source of information for the non-verbal communication concerning health. The detection of abnormalities from the human face is an emerging research domain as it helps to enhance the pain management of the non-communicative patients. For the automatic pain intensity estimation, several deep learning methodologies are available, still the estimation with enhanced accuracy is a challenging task. This research applies, reviews, and analyzes various deep learning methods for pain intensity estimation and identifies the research gaps in obtaining a more accurate deep learning method for real-time application processing. In this work, an experimental pain detection model was tested with deep learning methods and the performance was evaluated using evaluation metrics. The highest accuracy achieved is 74%, sensitivity is 79.5% and specificity is 72.6% with BiLSTM method. The research suggests an optimizer with ensemble approach will be more accurate for the automatic pain intensity estimation.