IOU and mAP

This tutorial will explain about IOU and mAP as evaluation index.

How to evaluate model

After pressing "RUN", the training state appears in dashboard.

In the box of "Learning Curve", the state of training is shown as learning curve. Learning curve indicates each learning curve for "Train" and "Validation". The X axis is epoch, and the Y axis is validation loss. The more the curve gets close to 0, the more training goes well.

The box in the right side shows model performance.

IOU and mAP

IOU is the abbreviation of Intersection Over Union, which indicates the certain of overlap of the number of predicted bounding box and the correct number of bounding box. The higher the value is, the higher the accuracy is.

mAP is the abbreviation of mean Average Precision, which indicates whether detected object is correct. The higher the value is, the higher the accuracy is.

Evaluation using test data

When scrolling down, there appears "Prediction Sample", which is the prediction result for "prediction_dataset".