How to evaluate the performance of the model on test images?

[DikshitV]

@AlexeyAB @pjreddie

I am working on custom Dataset using Yolov4-tiny custom. I have finished the training and saved the weights. Then used this weights for detection on batch of test images using the below command.

!./darknet detector test data/obj.data cfg/yolov4-tiny-custom.cfg /mydrive/yolov4-tiny/training/yolo_custom_best.weights -thresh 0.6 -dont_show -ext_output < data/test_images.txt > result.txt

In some of the outputs, detector has failed to detect the object eventhough object exists and remaining images it has detected well.

How to evaluate the performance of the model based on batch of test images (how many are detected, how many are not detected even though object exists)? can we get confusion matrix for batch of test images?