Identifying and categorizing plant diseases in a timely manner are essential in maintaining sustainable agriculture and increasing crop production. This research adopts a powerful computing paradigm known as Deep Learning to address the challenges in computer vision associated with multiclass plant leaf disease classification. We have used the most recent deep learning architectures, namely ResNet50, ResNet101, and ResNet152 and our proposed hybrid model (Resnet-50 + Resnet-101 + Resnet-152 + U-Net++) in our work to classify plant diseases. The multiclass classification is achieved through the activation function of the sigmoid present in the output layer of the ResNet. Additionally, the U-Net++ architecture is employed for segmentation tasks. The dataset utilized comprises multiple distinct classes. To enhance the efficiency of our models, transfer learning is applied, and the previously trained models are adjusted to fit the complexities of the plant disease categorization and segmentation domains. Upon thorough evaluation, the results reveal the promising performance of particular ResNet architectures (ResNet50, ResNet101, ResNet152) with U-Net++ and our proposed hybrid model (Resnet-50 + Resnet-101 + Resnet-152 + U-Net++) achieving 95.46, 93.96, 95.03, and 98.7% validation accuracies, respectively. This research contributes valuable insights into the utilization of deep learning techniques, particularly ResNet architectures and U-Net++, for accurate and efficient multiclass classification and segmentation of plant diseases.