Gender detection of silkworm pupae in silk seed production process is an important aspect for quality silk seed production and usually done manually with experts and needs to be automate in error free mode. However, before attempting this, it is very important to understand the importance of the dorsal and ventral surfaces of silkworm pupae for an accurate classification of gender. Differentiating between the dorsal and ventral surfaces of silkworm pupae is very important to pursue an accurate gender classification, which is one of the preliminary steps in the gender detection processes in the silk seed production process. Any slight inaccuracy during the surface determination process would result in an error in the gender classification, and this can adversely affect the silk seed production process. Some features that are unique distinctive features for gender detection are located mainly on the ventral surface of the silkworm pupae. Thus, it becomes important to ensure that there is an accurate distinction between the dorsal and ventral surfaces to prevent such classification biases in the silk seed production process. In the present study, using CNN, our objective is to classify the dorsal and ventral surfaces of silkworm pupae. The model implements K-fold validation and achieves a mean accuracy of 99. 70% and a standard deviation of 0.0040 with a perfect classification performance (100% precision) in several test cases that indicate the robustness and efficiency of the model. The proposed method aims to be an initial step in the integration of surface recognition in gender classification systems that improves reliability and productivity in silk seed production processes.