Metc2STM: Metachrosis Coordinate Optimized Convolutional Bidirectional Network for Video Captioning

Video captioning is capturing attention in the computer vision domain. Generating descriptions for video clips is a complex task and the prevailing techniques create challenges like computational issues, generating incorrect captions, the requirement of large databases, and so on. To conquer such limits, this study presents a Metachrosis coordinate optimized Convolutional bidirectional network (MetC2STM) model for accurate caption generation integrating the synergic strength of both the Bidirectional Long Short-Term Memory and Convolutional neural network. Specifically, the incorporation of Metachrosis coordinate hunt optimization algorithm enhances the tuning ability of the model. An encoder is aided to obtain the characteristic maps from the video frames and an embedded network is deployed for text embedding. In addition, the applied non-local means de-noising enhances the quality of the input frames. Furthermore, the proposed MetC2STM model achieved improvements in BLEU with 46.61, a METEOR of 25.37, and ROUGE scores of 69.26 respectively