Acute Lymphoblastic Leukemia (ALL) is a hematologic cancer where the lymphoblasts in the bone marrow are proliferating uncontrollably. It is one among the lethal cancer especially in pediatric population. The conventional methods of diagnosis of ALL are subjective and can sometimes miss to detect the early stages of ALL. Early diagnosis of ALL is therefore essential for appropriate treatment of the affected individuals. Now a days various computer vision based Artificial Intelligence models are developed for the robust detection of ALL. In this study, we utilized the CNMC 2019 dataset to enhance the accuracy and robustness of ALL detection aiming to provide valuable diagnostic insights into the disease. To address the data imbalance strong augmentation technique (SMOTE) and weak augmentation technique is used. A Novel ensemble-based Deep Learning model was designed leveraging two pre-trained model architecture that can detect ALL in an effective way. Deep feature extraction from individual models followed by feature fusion is incorporated to achieve the best results. The developed framework yielded an accuracy of 96.89% and F1-score of 97.71% on test data. In conclusion, the proposed model demonstrates reliable efficacy in detecting Acute Lymphoblastic Leukemia (ALL).