Inputs are very first handed through some completely linked layer, to the double-layer residual multihead focus as shown in Fig. seven. Residual networks (Kaiming He, 2016), include feedforward to prevent neurons from enduring exploding or vanishing gradients during the educational approach. The thoroughly related levels inside the residual block (