The increasing global burden of cancer demands timely and accurate diagnostic tools to improve patient survival rates and support clinical decision-making. This research introduces an intelligent deep learning framework designed to automatically detect multiple cancer types—including brain, breast, lung, cervical, colon, kidney, oral, and lymphoma—through analysis of medical images. The system is built using Convolutional Neural Networks (CNNs), with a focus on VGG and EfficientNet architectures, which are known for their robust performance in image classification tasks. A diverse set of cancer-related medical images is collected and pre-processed through normalization, resizing, and augmentation techniques to enhance the dataset's quality and variability. Both CNN models are trained on this data and evaluated using performance indicators such as accuracy, recall, and precision. The results reveal that the VGG architecture consistently achieves higher accuracy and computational stability compared to EfficientNet across various cancer classes. Based on this performance, the VGG model is deployed in a web-based interface, developed using the Flask framework, which allows users to upload images and receive real-time predictions. The system aims to assist medical professionals by reducing the time and potential human errors involved in manual diagnosis. Overall, the study demonstrates that CNN-based models, particularly those using the VGG architecture, have strong potential to support early detection of cancers and contribute to better healthcare outcomes.