Strawberry leaves are affected by fungal diseases by regarding the combination of features concerning about the absence of pathogens and the ideal atmosphere. Even though the fungal diseases occur often in their life cycle, they can weaken entire plants without causing any major impact. In this research a novel STR-REGNET has been suggested for the detection of strawberry leaf disease. The Adaptive Unsharp Mask Guided Filter (AUMGF), which reduces noise and improves key features including textures and edges, is used to preprocess the leaf pictures. The proposed model begins by an image of strawberry leaves, indicating both healthy and unhealthy leaves. An extremely effective Convolutional Neural Network (CNN) architecture called RegNet is utilized for feature extraction. It is used to modify an unprocessed image of strawberry leaves into a feature set that provides information. To transform into features, it makes advantage of pooling and effective convolutional blocks. Three distinct classifications are identified by the performance metrics of the strawberry leaf disease detection image: normal, powdery mildew, and leafspot. The collected dataset shows that the proposed STR-REGNET model obtains an overall accuracy of 95.85%. The proposed RegNet model's accuracy was greater than that of DenseNet, ResNet, and ResNest by 4.52%, 5.61%, 3.14%, 17.02%, and 8.9%, respectively. The proposed STR-REGNET model is better than CNNs, VGG 16, CNN, and SVM in terms of total ACC, obtaining 2.37%, 6.5%, 0.05%, and 18.17%, respectively