Subretinal injection (SRI) is a delicate procedure that requires precise cannula placement to ensure successful drug delivery beneath the retina surface. This paper presents the implementation of a convolutional neural network (CNN) for assessing the placement of a cannula relative to the injection point in SRI procedures. The proposed approach processes intraoperative images of the insertion process and classifies the feasibility of injecting fluid to form a bleb. A comparative analysis is performed to evaluate the CNN’s performance under various conditions, identifying the optimal configuration for real-world applications. The results demonstrate the potential of our approach to enhance the accuracy and safety of SRI procedures through automated assessment.