Estimating Maximum Likelihood using the Combined Linear and Nonlinear Function in NMS for Object Detection
D. Hema, Sathya Kannan · 2020
In object detection, Non-Maximum Suppression helps to eliminate the bounding boxes that fall below the minimum confidence threshold among N number of predicted bounding boxes. There are instances where the returned bounding boxes might be a false positive one or it may return multiple bounding boxes. To identify one accurate bounding box prediction, an algorithm that includes truer positive and reject false positives should be designed for object detection. To address this issue, many IOU (Intersection over Union) guided NMS algorithm has been suggested. The proposed research work combines linear and nonlinear functions to implement a novel fuzzy IOU-guided NMS to estimate the maximum likelihood (i.e., bounding box in our case) to fine-tune the task of object detection. The proposed work reveals that the predicted bounding boxes are very close to the ground truth and is efficient in eliminating false positives among all predicted bounding boxes.