How can I find the Average Precision and Average Recall from the eval.py file?

[AtriSaxena]

I want to find the Average Recall and Precision values. I have tried, printing these value Code link
rec = tp / float(npos) prec = tp / np.maximum(tp + fp, np.finfo(np.float64).eps)

But it is printing long list of values which i am not able to get.

[jiawen7777]

I meet the same problem, have you solved it? Thank you very much

[AtriSaxena]

I have solved but it was long time ago. I don't remember but tell me more about the problem you are getting? I have used custom data in VOC format.
Can you show me error you are getting?

[wallcuber]

I have the same question. The rec and prec is so long. How did you deal with this？

I have solved but it was long time ago. I don't remember but tell me more about the problem you are getting? I have used custom data in VOC format.
Can you show me error you are getting?

[wallcuber]

I have solved but it was long time ago. I don't remember but tell me more about the problem you are getting? I have used custom data in VOC format.
Can you show me error you are getting?

Does rec[-1] and prec[-1] are the final average Recall and Precision ?

[AtriSaxena]

After this Link I have added this code.

rec += [rec] prec += [prec]

And also define rec = [] and prec = []

Than you can print
print('Average Recall = {:.4f}'.format(np.mean(rec))) print('Average Precision = {:.4f}'.format(np.mean(prec)))

        filename = get_voc_results_file_template(set_type, cls)

        rec, prec, ap = voc_eval(
           filename, annopath, imgsetpath.format(set_type), cls, cachedir,
           ovthresh=0.5, use_07_metric=use_07_metric)
        aps += [ap]
        rec += [rec]
        prec += [prec]
        print('AP for {} = {:.4f}'.format(cls, ap))
        with open(os.path.join(output_dir, cls + '_pr.pkl'), 'wb') as f:
            pickle.dump({'rec': rec, 'prec': prec, 'ap': ap}, f)
    print('Mean AP = {:.4f}'.format(np.mean(aps)))
    print('Average Recall = {:.4f}'.format(np.mean(rec)))
    print('Average Precision = {:.4f}'.format(np.mean(prec)))```

The whole loop will be something like this. 
If you are using Custom dataset than you need more changes. 
If this solves tells me.

[jiawen7777]

wow！thank you! It works!

…

---Original--- From: "ATRI SAXENA"<notifications@github.com&gt; Date: Mon, Apr 27, 2020 02:18 AM To: "amdegroot/ssd.pytorch"<ssd.pytorch@noreply.github.com&gt;; Cc: "Comment"<comment@noreply.github.com&gt;;"skylineJLU"<1457013276@qq.com&gt;; Subject: Re: [amdegroot/ssd.pytorch] How can I find the Average Precision and Average Recall from the eval.py file? (#352) After this Link I have added this code. rec += [rec] prec += [prec] And also define rec = [] and prec = [] Than you can print print('Average Recall = {:.4f}'.format(np.mean(rec))) print('Average Precision = {:.4f}'.format(np.mean(prec))) `for i, cls in enumerate(labelmap): filename = get_voc_results_file_template(set_type, cls) rec, prec, ap = voc_eval( filename, annopath, imgsetpath.format(set_type), cls, cachedir, ovthresh=0.5, use_07_metric=use_07_metric) aps += [ap] rec += [rec] prec += [prec] print('AP for {} = {:.4f}'.format(cls, ap)) with open(os.path.join(output_dir, cls + '_pr.pkl'), 'wb') as f: pickle.dump({'rec': rec, 'prec': prec, 'ap': ap}, f) print('Mean AP = {:.4f}'.format(np.mean(aps))) print('Average Recall = {:.4f}'.format(np.mean(rec))) print('Average Precision = {:.4f}'.format(np.mean(prec)))` The whole loop will be something like this. If you are using Custom dataset than you need more changes. If this solves tells me. — You are receiving this because you commented. Reply to this email directly, view it on GitHub, or unsubscribe.

[jiawen7777]

I think I might miss your reply before
I just saw the reply this morning, Sorry

[AtriSaxena]

Yeah, I replied last night. I was little busy. Great, it helped you.

[zyx961022]

ValueError: non-broadcastable output operand with shape (11725,) doesn't match the broadcast shape (1,11725)

Why do I report this error and how to solve it? please help me， thank you！

[decoli]

ValueError: non-broadcastable output operand with shape (11725,) doesn't match the broadcast shape (1,11725)

Why do I report this error and how to solve it? please help me， thank you！

Have you solved the problem? I encountered the same problem as you.
Then I referred aps += [ap] and @AtriSaxena 's comment, I added these codes:

...
recs = []
precs = []
...
        ...
        recs += [rec]
        precs += [prec]
        ...
    ...
    print('Average Recall = {:.4f}'.format(np.mean(recs)))
    print('Average Precision = {:.4f}'.format(np.mean(precs)))
    ...

This will work, but I’m not sure if it’s right. What do you think?