As a significant health concern worldwide, Laryngeal cancer detection is critical for early diagnosis and effective treatment. While keeping this in mind, we introduce image classification to detect the initial cancer potentials. For the detection, we have worked on the hyperparameter tuning and modifications of our models to ensure the best results from the models over our dataset. Data modification and augmentation schemes help the hyper-tuned models to perform better in this classification of infected and normal classes. Throughout this research, we tried to classify the classes with different types of CNN models like VGG16, VGG19, ResNet50, InceptionV3, DenseNet121, and MobileNetV2 to compare and get the best overall results. As discussed earlier, our research was based on two classes of images labeled as infected and normal. While the hyper-tuned VGG16 model with data augmentation outperforms others with a training accuracy of 99.2% and a test or validation accuracy of 99.13%.