Volumetric additive manufacturing (VAM) has emerged as a promising strategy for fabricating biomedical devices owing to its capacity to produce complex architectures in a single step. This technique requires precise spatial regulation of light dose and meticulous control over exposure time. However, conventional exposure timing methods still rely heavily on operator subjectivity, leading to labour-intensive workflows and inconsistent printing reproducibility. To address these challenges, this study introduces an integrated volumetric printing system featuring in-situ monitoring functionality, coupled with an intelligent visual inspection framework that combines schlieren imaging with deep learning. Through systematic optimisation of the viscosity of refractive index-matching solutions and background illumination parameters, substantial enhancements were achieved in both quality and stability of real-time imaging, enabling the acquisition of extensive image datasets throughout the fabrication process. A convolutional neural network based on ResNet18 was then employed to accurately recognize and classify different printing stages. Experimental results indicate that when using a refractive index-matching solution with a viscosity below 650 mPa s under 1000 lx background illuminance, the system captures high-contrast, low-noise schlieren images that clearly elucidate dynamic solidification behaviour. The integration of deep learning further underscores its potential for AI-driven interpretation and quality assurance in VAM. These outcomes provide crucial technical foundations for real-time monitoring and closed-loop control in VAM, underscoring new opportunities in precision manufacturing and biomedical applications.