Responsive polymers can alter their properties or shape according to stimuli such as stress, light, among others. The stimuli may be strategically programmed at assigned triggering points in a structure using additive manufacturing, which enables controlled deformation that is crucial for reversible shape transformation. Understanding and controlling shape transformation holds significance in topology optimization, or generative design, among other applications, where a material may not be reduced by machining; it can instead be reversibly transformed to perform evolving functions. However, handling responsive polymers can be challenging, as they react to various stimuli. The present research introduces a contactless technique using deep classification to detect and measure shape changes in responsive polymers. The application of deep network in responsive structural element (characteristic points) detection has not been attempted before. The classification categorizes polymers according to the degree of shape change, which aids in the verification of selectively programmed stimulus-responsive properties. By correlating the output with trained data, accurate classifications were made regarding the degree of shape change. The shape changes were then compared with different deformation angles, lengths, or heights through a novel formula-based correlation. The classification helps reconfigure the material if the deformation is not within the desired range. The classification and comparison were performed with AlexNet, which possesses a simple architecture. An accuracy of 100% was achieved by fine-tuning the hyperparameters of the same network. The present work thus shows the possibility of combining multiple tasks more efficiently with less data and computational resources using simple network.