Evaluation of Small Object Detection in Scarcity of Data in the Dataset Using Yolov7

Object detection had gained importance in previous decade due to large amount of data that is being generated throughout the world by cameras, mobile phones, satellite imaginary, medical image, social media, UAV etc. As hardware cost to render these images had been reduced significantly and we have access to plethora of algorithms, framework to detect the object and use this information to solve day to day problems. The object detection is most researched area but it still fails to detect and recognize small objects as detecting large objects had got more focus. But small object detection had got less attention and the algorithms and methodology developed for detecting large object does not yield the desired results and accuracy. In this paper we attempt to detect small objects by using state of art algorithm yolov7 and roboflow and try to evaluate the robustness of object detection with scarcity of data in dataset.