This paper presents a novel deep learning model called MAPCRCI-DMPLC for the early detection of diabetic retinopathy (DR). The model is designed to classify retinal fundus images into five stages of DR with high accuracy and minimal time consumption. The model incorporates preprocessing, ROI extraction, feature extraction, and classification steps. The preprocessing step uses a MAP estimated local region filtering technique to enhance image quality and reduce noise. ROI extraction is performed using Camargo's index to identify infected regions. Texture features are extracted using Concordance Correlative Regression, while color features are extracted from the image. The classification is performed using the swish activation function to achieve high accuracy and low error rates. The model is evaluated using metrics such as peak signal-to-noise ratio (PSNR), disease detection accuracy (DDA), false-positive rate (FPR), and disease detection time (DDT). The results show that the proposed model outperforms existing state-of-the-art approaches in terms of accuracy, FPR, and time complexity. The model's performance is validated using a retinal image dataset, and it demonstrates significant improvements in disease detection accuracy and reduced error rates compared to conventional methods. The model's innovative approach to preprocessing, feature extraction, and classification contributes to its effectiveness in early DR detection.This paper presents a novel deep learning model called MAPCRCI-DMPLC for the early detection of diabetic retinopathy (DR). The model is designed to classify retinal fundus images into five stages of DR with high accuracy and minimal time consumption. The model incorporates preprocessing, ROI extraction, feature extraction, and classification steps. The preprocessing step uses a MAP estimated local region filtering technique to enhance image quality and reduce noise. ROI extraction is performed using Camargo's index to identify infected regions. Texture features are extracted using Concordance Correlative Regression, while color features are extracted from the image. The classification is performed using the swish activation function to achieve high accuracy and low error rates. The model is evaluated using metrics such as peak signal-to-noise ratio (PSNR), disease detection accuracy (DDA), false-positive rate (FPR), and disease detection time (DDT). The results show that the proposed model outperforms existing state-of-the-art approaches in terms of accuracy, FPR, and time complexity. The model's performance is validated using a retinal image dataset, and it demonstrates significant improvements in disease detection accuracy and reduced error rates compared to conventional methods. The model's innovative approach to preprocessing, feature extraction, and classification contributes to its effectiveness in early DR detection.