In this work, we introduce the topological data analysis layer that estimates persistent homology on attributes extracted from convolutional layers for image classification. This method shows that topological information can be utilized to upgrade network performance. This work focuses on applying persistent images on the deep convolutional layer to learn topological features and also exploring the behavior of topological data analysis on various convolutional neural network architectures like sequential architecture and extended width architecture. Based on our empirical analysis, we exhibit the significance of topological data analysis on convolutional neural networks by attaining reliable scores on classification tasks on benchmark datasets.