Confusion Matrix

[altimis]

Confusion Matrix

[vishnuvardhan58]

Hello @Altimis ,it is working if I have more images. Thankyou.

[Altimis]

Great! Happy to help. I will fixe this issue as soon as possible, thank you .

[Altimis]

Hello @vishnuvardhan58 , could you please show me the whole error message. Did you generate the gt_tot and pred_tot lists of ground-truth and predictons ?

[vishnuvardhan58]

Hello @Altimis ,here is the whole error message

[vishnuvardhan58]

yes , I have generated gt_tot and pred_tot lists of ground-truth and predictons .

[Altimis]

Hi again @vishnuvardhan58 , I think that I spotted where is the problem. The porblem is in the line
columns = range(1, len(np.unique(y_test))+1)
Either you need to remove the "+1" or set len(np.unique(y_test)) to the number of your classes + 1 (background). If neither f these work, please let me know.

[Altimis]

I would like to take a look at your gt_tot and pred_tot

[vishnuvardhan58]

Hello @Altimis , I am still getting the same error after making the above changes ,

Please find the gt_tot and pred_tot values , I have 3 images in my test_set

[Altimis]

Ah okey I get it. In fact, the columns = range(1, len(np.unique(y_test))+1) means that there is (1,5) since len(np.unique(gt_tot)) = 4 in your case. But your pred_tot vector doesn't contain background class (0) so it's len is equal to 3 and not 4 like gt_tot.
The problem is exacty in df_cm = DataFrame(confm, index=columns, columns=columns) since I've set index = columns = range(1, len(np.unique(y_test))+1) since confusion matrix should be square. In fact i didnt include in my code the case where the model doesn't detect background intead of an object (you used only 3 images, if you used more for testing I'm sure this probleme wont occure.

[Altimis]

I will try to solve this problem by including the case where the model doesn't miss ^^. For now, you can test your model on more data and it will probably work

[vishnuvardhan58]

I will try with more data @Altimis , and I will let you know.