Optimizing cannabis crop yield and quality necessitates accurate, automated leaf disease classi-fication systems for timely detection and intervention. Existing automated solutions, however, are insufficiently tailored to the specific challenges of cannabis disease identification, struggling with robustness across varied environmental conditions and plant growth stages. This paper introduces a novel Hybrid Adaptive Multi-Intelligence System for Deep Learning Ensembles (HyAMIS-DLE), utilizing a comprehensive dataset reflective of the diversity in cannabis leaf diseases and their progression. Our approach combines non-population-based decision fusion in image prepro-cessing with population-based decision fusion in classification, employing multiple CNN archi-tectures. This integration facilitates a significant improvement in performance metrics: Hy-AMIS-DLE achieves an accuracy of 99.58 %, outperforming conventional models by up to 4.16 %, and exhibits superior robustness and an enhanced Area Under the Curve (AUC) score, effectively distinguishing between healthy and diseased leaves. The successful deployment of HyAMIS-DLE within our Automated Cannabis Leaf Disease Classification System (A-CLDC-S) demonstrates its practical value, contributing to increased crop yields, reduced losses, and the promotion of sus-tainable agricultural practices.