graph LR
Post_processing_Orchestrator["Post-processing Orchestrator"]
Intersection_Over_Union_Calculator["Intersection Over Union Calculator"]
Post_processing_Orchestrator -- "calls" --> Intersection_Over_Union_Calculator
The core of this subsystem focuses on refining raw object detection outputs into a set of precise and distinct bounding box predictions. The Post-processing Orchestrator serves as the central control, taking the initial predictions from the neural network. It first filters out low-confidence detections and then orchestrates the Non-Maximum Suppression (NMS) process. A critical dependency for NMS is the Intersection Over Union Calculator, which the Post-processing Orchestrator repeatedly calls to quantify the overlap between bounding boxes. This interaction ensures that only the most accurate and non-redundant detections are ultimately presented, forming a streamlined and effective post-detection pipeline.
This component acts as the primary controller for the post-processing pipeline. It receives raw bounding box predictions and confidence scores from the neural network. Its core responsibilities include applying a confidence threshold to filter out low-probability detections and executing the Non-Maximum Suppression (NMS) algorithm to eliminate redundant, overlapping bounding boxes, ensuring only the most confident and distinct detections are retained.
Related Classes/Methods:
This is a utility component dedicated to computing the Intersection Over Union (IOU) metric between two given bounding boxes. IOU is a crucial measure of overlap and is fundamental for the Non-Maximum Suppression (NMS) algorithm, allowing the system to quantify how much two predicted boxes overlap.
Related Classes/Methods: