Detection of emotion from various aspect is the most important thing in social media environment. Numerous research has gone into giving robots the ability to recognize emotions. Most prior computer vision attempts focused on assessing facial expressions and, in certain circumstances, body position. This strategy works nicely in certain situations. Their performance is restricted in natural settings. Studies suggest that the scene setting, together with facial expression and body stance, helps us perceive people's emotions. However, owing to a paucity of data, the context processing for automated emotion identification has not been fully studied. In this paper we proposed a system for detection of emotion using real time visual and context-based features. The major objective of this research to identify the sentiment from large video based on context. To achieve this functionality, it needs to extract numerous frames from video and validate all frames using proposed algorithms. This research basically carried out the extraction of hybrid features from real time image as well as video dataset and build classifier to selective features. The hybrid deep learning classification algorithm has used for predict the correct sentiment with Hybrid Convolutional Neural Network (H-CNN) and Recurrent Neural Network (RNN). In extensive experimental analysis evaluation has done in terms of accuracy which obtains better than traditional machine learning and deep learning classification algorithms.