A significant amount of crop yield is being lost due to insect pests attack annually. The problem becomes sever in developing agrarian countries like Ethiopia. Insect identification at larva stage is essential because easy management. due to its less movability than starting flying in the next stage. However, high similarities among armyworm insects at the larvae stage makes crop armyworm insect identification is the most challenge task for agricultural crop experts. These in turn hindered the mitigation measures. This study aims to develop an armyworm insect detection and classification model which simplifies the recognition tasks of agricultural crop experts. The model has been developed using a combined approach of image processing and pattern recognition techniques. The designed armyworm insect recognition model follows image processing techniques such as image preprocessing, image analysis and image understanding. In image preprocessing tasks, image size normalization, image denoising, segmentation, and RGB to gray scale conversion tasks were conducted. Immediately, after preprocessing appropriate CNN model were designated thereby inspiring the VGGnet model and restructuring the CNN model thereby inserting different filter size, batch normalization and dropout layer. Finally, image understanding was done by examining CNN-Softmax classifiers. We, therefore, examined the performance of the CNN for both feature extraction and feature classification techniques. Better recognition performance was registered while considering the grayscale image of armyworm as an input to the CNN model (88.1%) RGB armyworm insects are used as an input to the CNN model (86.4%). We, therefore, recommended farther investigation of CNN model by restructuring the CNN layers with hidden layer by using better dataset size and needs to integrate increasing cleaning the dataset by continuing to collect more from the field.