High Efficiency Video Coding. (HEVC) is the product of a large collaborative effort from industry and academic community and reflects the new international standardization for digital video coding technology. Compression capability is the main goal behind the digital video compression technology. HEVC achieves this goal at the expense of dramatically increasing coding complexity. One such area of increased complexity is due to the use of a recursive quad-tree to partition every frame to various block sizes, a process called prediction mode. This exhaustive prediction process calculates a rate-distortion metric for each block in order to choose the best mode. This paper proposes three Convolutional Neural Networks based on VGGNet [1] for predicting the coding structure of HEVC quad-tree decision in order to minimize the coding complexity of HEVC at a cost of minimum loss of compression and image quality. As a result, our method reduces 73.69% encoding time with 1.10% of compression loss measured by BD-BR and no significant loss in PSNR.