IoU (Intersection over Union) Visualizer

This interactive demo helps understand how IoU works in object detection tasks. Draw ground truth (black) and prediction (colored) boxes to see how well they match. Double-click any box to delete it. The color of prediction boxes indicates the quality of the match:

  • Red: No significant overlap
  • Yellow: Overlap exists but below threshold
  • Green: Good match (above IoU threshold)

Debug Info


                

            


            
            

                

                    
                    
                    
                    
                

                

                    
                

            


            
            

                Demo image

                
            


            
            

                
Statistics

                

                    

                        
Ground Truth boxes: 

                        
Prediction boxes: 

                    

                    

                        
True Positives: 

                        
False Positives: 

                    

                

                

                    

                        See how these results translate into a confusion matrix. False negatives (FN)
                        are ground truth boxes without a matching prediction. True negatives (TN) in object
                        detection would technically be all possible box locations where there isn't an object
                        and we correctly didn't predict one - this number would be extremely large and not
                        meaningful, which is why object detection typically uses other metrics like
                        Precision and Recall instead of accuracy.
                    

                    
                        View Confusion Matrix