The Channel estimation in Vehicle-to-Vehicle (V2V) communications is critical due to the time-varying characteristics of the V2V channel. Hence, the authors propose a hybrid approach towards channel estimation by combining Pilot Symbol Aided Channel Estimation (PSACE), Improved Crow Search Algorithm (ICSA), and Decision Directed Channel Estimation (DDCE) by employing Weight Regularized Convolutional Neural Network (WCNN). Hybrid channel estimation by combining the properties of PSACE and DDCE. PSACE suffers from pilot overhead issue due to the demand of huge number of pilots for channel estimation. On the other hand, DDCE suffers from error propagation in fast fading environments. The proposed system employs an ICSA and WCNN to overcome the issues faced by channel estimation methods. Further, the proposed strategy replaces the analytical modeling of the channel with the help of the neural network. Optimized pilots obtained from ICSA along with data symbols are fed to WCNN to estimate the channel in a non-stationary environment. Finally, the performance of the proposed method is evaluated in terms of Bit Error Rate, Mean Square Error, and Packet Error Rate and the simulations are carried out under various V2V scenarios for testing its sturdiness.