Accurate, real-time pressure monitoring is critical for the effective management of chronic venous disease, necessitating advanced solutions in compression therapy. This study presents a novel smart compression stocking system that integrates cost-effective, laboratory-fabricated textile capacitive pressure sensors within compression garments, a compact edge control unit, a dedicated mobile application, and advanced edge signal processing algorithms. Although simply fabricated textile-based sensors are highly sensitive and scalable, their accuracy can be affected by parasitic capacitance, electromagnetic interference, proximity effects, and deformation on curved surfaces. To address these challenges, we developed a physics-informed encoder–decoder deep neural network architecture that enhances noise reduction and enables high-resolution pressure detection. The encoder, constructed from the symmetrical stacked autoencoder, performs nonlinear feature extraction and incorporates soft physical constraints, while the decoders are tailored for specific downstream tasks. This lightweight, interpretable architecture is computationally efficient and suitable for deployment on resource-constrained edge devices. Empirical validation was conducted using representative datasets, comprising 22,498 capacitance–pressure data pairs from 100-minute flat surface pressure recordings and 75,888 pairs from 12-minute curved surface pressure recordings over ten days. The proposed algorithms achieved a root mean square error (RMSE) of 0.7891 mmHg for flat surfaces, a 38% improvement over polynomial regression, and RMSEs of 0.0283 mmHg, 0.0633 mmHg, and 0.0387 mmHg at curved surface positions B, B1, and C, representing improvements of 98%, 97%, and 98%, respectively. These results demonstrate the system’s potential to deliver precise, real-time, accessible, and scalable pressure monitoring, marking a significant advancement in the conservative management of chronic venous disease.