Abstract: Insects are the main biotic stress responsible for crop losses in the world. To solve this, the Forestry and Agricultural Biotechnology Institute (FABI) needs assistance in identifying and classifying insect pests using a tiny dataset containing the Sirex and Gonipterus insects. We annotated the full FABI dataset with bounding boxes and different life stages categories. Furthermore, transfer learning techniques and a publicly available pest dataset, IP102, were utilized to finetune different variants of the residual and mobile neural networks models. Data augmentation techniques and image enhancement techniques were used to increase the dataset and address the issue of low light images. The best performing models developed in this study can be used as backbones for object detection algorithms.