0
$\begingroup$

I am training some Object-Detection-Models from the TensorFlow Object Detection API and got from the evaluation with MS COCO metrics the following results for Average Precision:

IoU = 0.5;0.9 maxDets = 100 area = small AP = -1.000

The other values all make sense to me. But I don´t know what the -1.000 stands for. Does it mean that there are no small objects in my dataset to be detected?

$\endgroup$

2 Answers 2

0
$\begingroup$

APsmall stands for Average Precision for small objects.


This website contains all the descriptions about the metrics: https://cocodataset.org/#detection-eval

If you want to know more about the metrics evaluation (how to calculate this metric), you can find the metric's source code here: https://github.com/cocodataset/cocoapi/blob/master/PythonAPI/pycocotools/cocoeval.py

$\endgroup$
0
$\begingroup$

Does it mean that there are no small objects in my dataset to be detected?

Yes.

From the evaluation source code as mentioned by the other answer, L334:

-1 for the precision of absent categories

From the coco dataset website, Detection Evaluation:

$AP_{small}$: AP for small objects: area < $32^{2}$
$AP_{medium}$: AP for medium objects: $32^{2}$ < area < $96^{2}$
$AP_{large}$: AP for large objects: area > $96^{2}$

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .