Accurate classification of colorectal cancer histology images plays a pivotal role in diagnosis and treatment planning. In this article, the efficacy of utilizing the EfficientNetB3 architecture with sigmoid activation function and the Adamax optimizer for histopathological image classification had been investigated. Several data preprocessing techniques, like the CLAHE technique and Weiner filter methods, are used, which eliminate unwanted artifacts, resulting in significant enhancements in image quality. By integrating context information into the encoder-decoder networks. With an emphasis on achieving high accuracy, we leverage the sigmoid activation function and the Adamax optimizer, tailored to suit the characteristics of the dataset entitled Kather-texture-2016-image comprising 5000 histology images. The dataset, comprising colon adenocarcinoma and benign tissue classes, serves as a novel input for training, capturing intricate histological nuances. The current research addresses the challenges inherent in categorizing diverse tissue types and variations in cellular characteristics within colorectal cancer histology images. Through extensive experimentation, a classification accuracy of 97.8% was achieved, demonstrating the effectiveness of the proposed approach. Leveraging deep learning techniques, a differentiation between healthy and diseased large intestine cells had been achieved, contributing to the advancement of diagnostic accuracy in colorectal cancer detection. The comparative analysis conducted on a dataset of 5000 images showcases promising results, indicating the potential of the proposed approach to enhance the accuracy of histopathological image identification. The findings of this study hold significant implications for improving diagnostic practices in colorectal cancer detection.