Future Frame Predictor for Oceanography

In the fields of Marine exploration, disaster relief and ecosystem monitoring, it is very much important to find out the underwater events. Regular technqiues mostly depend upon sensors and other hardware devices.The issue with sensors and other hardware components is that it is hard to maintain since it can easily get damaged or spoiled. Also high level of maintanence is required which also result in high cost.So in this study, we provide a unique AI-driven method that is powered by deep learning and image based analysis which serve as a replacement in the place of actual hardware sensors. Our approach also helps us to predict the next event that is going to happen underwater.This is achieved using FutureGAN.Also to describe the future events in a textual format, we use an VLM model known as PaLI-GEMMA which is used to describe the image.We also use adversarial training, perceptual loss, and optical flow limitations to improve motion prediction. The suggested method's ability to provide the sensor outputs is demonstrated using various underwater datasets.This model takes a major step in the field of oceanography and marine studies.