More importantly, prediction of software defects provides us with the opportunity of detecting areas of the code which can cause bugs long before it is utilized by detecting them in advance. It thought how to predict software defects using CNN as a powerful deep learning technique by computing them using different code measures and representations. Differently to common approaches, CNNs enable automatic extraction of features in raw data such that persons are not forced to manually engineer them. The software program utilized in the current research converts software codes into either greylevel or sequence matrixes to enable the CNN model to peruse through them. When tested on standard datasets, the method outperformed traditional classifier on the basis of precision, recall, accuracy and F1-score. These findings imply that CNN-based models are very useful at learning the intricate patterns and logic in code, and this is the reason why such models are a highly deemed solution when it comes to predicting software defects automatically.