Not enough input arguments for plotting Precision-Recall curve

Greetings,

I want to evaluate my Yolov3 model by plotting the Precision-Recall curve. I tried to follow the example provided in the "Object Detection Using Yolo v3 Deep Learning' (https://www.mathworks.com/help/vision/ug/object-detection-using-yolo-v3-deep-learning.html). However the error displayed as follows:

Error using plot

Not enough input arguments.

Error in testyolo (line 41)

plot(recall, precision);

May I know what is the problem with my code? My code is

celldetector = load('trainedyolov3Detector-2022-11-04-00-30-56.mat');
testData = combine(imdsTest, bldsTest);
detector = celldetector.yolov3Detector;
results = detect(detector,testData,'MiniBatchSize',8);
[ap,recall, precision] = evaluateDetectionPrecision(results, testData);
figure;
plot(recall, precision);
xlabel('Recall')
ylabel('Precision')
grid on
title(sprintf('Average precision = %.1f', ap))

Answer 1

[Walter Roberson]

evaluateDetectionPrecision

"For a multiclass detector, recall and precision are cell arrays, where each cell contains the data points for each object class."

So it appears you have a multi-class situation, and your recall and precision are being returned as cell arrays. plot() cannot handle cell arrays

[Fahmi Akmal Dzulkifli]

Thank you Sir for you explanation. Is there any ways to solve it or have other analysis that can be applied besides calculate the loss function?

[Walter Roberson]

You could probably do something similar to

hold on
cellfun(@(R,P) plot(R,P), recall, precision)
hold off
xlim auto; ylim auto

It might not be exactly this, as I do not know exactly what are in the cell arrays. And you probably want to add some kind of legend, but I am not sure where the label information is stored in this situation.

(I do not have example multiclass data around to test this with.)

[Fahmi Akmal Dzulkifli]

Sir, I already tried with your code, but it seems the graph had two signals, which was different from the example provided in the example in ( https://www.mathworks.com/help/vision/ref/evaluatedetectionprecision.html ). May I know what is the different between these two codes

[Walter Roberson]

You would get one line for each class.