Both colorectal and pulmonary cancer have been serious conditions requiring immediate medical attention. With early detection and appropriate therapy, patients with certain tumors have experienced substantially better outcomes. People need to be aware of potential risk factors, symptoms, and recommended screening procedures that include their age and medical history. Initially, with the categorization of colorectal and pulmonary cancer, we employed three typical convolutional neural network (CNN) models in this paper: MobileNetV2, Inception-ResNetv2, and Xception. To these models, we have also applied an attention mechanism. We have selected these CNN-based learner models and created an ensemble model to upgrade the overall accuracy. The ensemble approach is referred to as MIXNet, where M is for MobileNetV2, I is for InceptionResNetv2, and X stands for Xception. We have utilized two datasets for experimental evaluation: the Chest CT scan images dataset and the freshly released LC25000 dataset. Using the suggested MIXNet model, we have obtained satisfactory results with accuracy scores of 99.50% and 98.15% on the LC25000 and Chest CT scan images datasets, respectively. This model has outperformed its constituent base models.