Incremental Learning for Object Detection of Unmanned Ariel Vehicles

Object detection is applied in many applications, such as unmanned aerial vehicles and autonomous vehicles, for multiple purposes due to its huge advancements.Object detection works as normal convolution neural network models with classification and localization for unmanned ariel vehicles.However, these traditional models of artificial intelligence suffer from catastrophic forgetting when trained on a new dataset, thereby compromising their ability to retain previously learned information.In this paper, we propose a novel integration of knowledge distillation with the famous YOLO object detection framework to overcome catastrophic forgetting.By using knowledge distillation, the model effectively transfers knowledge from previous tasks to new ones, preserving performance in earlier classes while retaining the previous information.The proposed framework is evaluated on Pascal VOC datasets in two classes to present the performance of incremental learning.Multiple experiments on these datasets suggest that our method has significantly improved the accuracy of previous classes in comparison to state-of-the-art methods.