Cyclones are one of the deadliest natural calamities capable of causing immense destruction. Knowing about natural disasters like cyclones in advancehelps in planning and preparations as they can beextremely dangerous. Numerous methods have beendeveloped in the past to track cyclones and gaugetheir severity after eye formation. The purpose ofthis paper is to track the movement of cyclones beforethe eye formation to avert cyclone-related damage ina fast and efficient way. This paper majorly focuseson cyclone forecasting before the formation of the eyewhich is a difficult task and demands effectiveness.The INSAT-3D satellite pictures consisting of IR andvisible images are analyzed and segmented using Detectron. Comparisons have been made between various alternate models, including CNN-LSTM and ahybrid model that combines CNN and BidirectionalLSTM. The model put forth in this paper is a combination of CNN and Bidirectional-GRU. The hybridmodel of CNN and Bidirectional GRU is trained onthe INSAT-3D dataset to estimate the next positionof the cyclone. The suggested model’s experimental results show an MSE of 1613.65 and an SSMI of(1.0,1.0).