Autonomous unmanned aerial vehicles (UAVs) are commonly used for wildlife exploration and animal monitoring. Therefore, bird attacks pose a significant challenge to UAVs. As we know, Traditional Bird Detection methods used for prevention against attacks may fail when the attacks occur from unobservable angles. However, the UAV can gain an early indication of an impending attack if it detects the location of the bird’s shadow and takes proactive measures to minimize the risk of damage. To address this, we present the ShadowBirdCUB dataset, derived from the CUB-200-2011 Dataset, which is used to train cutting-edge Deep Learning Algorithms for shadow detection. Experimental results using various Deep learning Object Detection (DLOD) models and performance metrics demonstrate promising effectiveness. Although this approach is limited to detecting attacks from certain angles, it is a valuable addition to existing bird detection methods.