In this study, we translate the topic of bird detection into a standard single keypoint detection problem. The model output a heatmap, which is an image composed of the probability values representing the centroids in each pixel. In addition to the heatmap, the model should also output the relevant properties for each center point, including the corresponding length and width of each center point to form the binding box
The experiments and the results have shown that our model is able to achieve the best speed-accuracy compared to the others.