<p>Massive or extensive approach of MIMO (Multiple Inputs Multiple Outputs) contains a high capability to attain a high rate of data and is one of the most preferred techniques to utilize the efficiency of channel feedback. Therefore, a feedback CSI mechanism based on deep learning in this paper Deep-EDM (Encoder Decoder Model) is proposed to ensure high-efficiency channel estimation with the least overhead CSI feedback. Additionally, the encoder and decoder are incorporated to analyze the low-dimensional depiction of different data structures. However, compression of CSI matrices at the side of the encoder as well as the CSI matrices being recovered are found at the decoder side. Furthermore, convolutional layers are used to obtain high feature quality and a completely connected layer is used for the compression of dimensions in the feedback CSI matrices. The efficiency of CSI feedback is improved by the use of CWC (Complex Weights coefficient)-CVNN (Complex Valued Neural Network) aka CWC-CVNN, this architecture utilizes the uplink as well as downlink medium magnitude correlation. Herein, the datasets of two different environments, for instance, outdoor and indoor cellular situations take into account the Cost 2100 database being used, and the cloud platform is used for simulation. A thorough study is performed, wherein the result of the Deep-EDM proposed model considering NMSE (Normalized Error Mean Square) and efficient correlation is compared to the traditional approaches of channel estimation. The results obtained manifest higher accuracy of channel estimation and enhanced spectral efficiency. Considering the proposed model, for indoor environments, the CSI compression accuracy improved by 16.7144%, 25.1578%, and 12.0512% at ratios of 1/4, 1/8, and 1/16 respectively over the previous ACRNet-20×. For outdoor environments, the gains were even more pronounced at 33.4551%, 42.98%, and 33.5691% for the same ratios compared to an existing model.</p>