search

dbolya / yolact

A simple, fully convolutional model for real-time instance segmentation.
MIT License
5.03k stars 1.32k forks source link

How to measure metrics? #181

Closed sainatarajan closed 5 years ago

sainatarajan commented 5 years ago

Hi Daniel. I have used your repo and the model performs very accurate predictions. But, I would like to calculate the values of metrics like iou_score, accuracy, f1-score, precision, and recall. Can you tell me whether these metrics are already available or we have to write custom code for metrics calculation?

dbolya commented 5 years ago

I have functions to compute some of those buried in the code (see box_iou and mask_iou in eval.py), but you'll likely have to implement most of it yourself. You can start by modifying calc_map and APDataObject's get_ap function in eval.py.

abhigoku10 commented 5 years ago

@dbolya any plans to include the ROC curve plotting with different thresholds in future

dbolya commented 5 years ago

@abhigoku10 It might be useful so sure, but it would be low priority.

sainatarajan commented 5 years ago

@dbolya Thanks for the information. I will try to use the existing code and if required modify them.

dbolya commented 5 years ago

@sainatarajan If you need anymore help let me know (by reopening this issue). I'll close the issue for now.

WalBouss commented 4 years ago

@dbolya Thanks for the information. I will try to use the existing code and if required modify them.

@sainatarajan have you implemented the ROC curve plotting or the AUC score? I so can you please share the code?

sainatarajan commented 4 years ago

@WalBouss I haven't plotted yet the curves. Maybe I could try it out this week.

kimhyuenjun commented 4 years ago

I am very interested in YOLACT's ROC curve.

How do you draw the ROC curve through the YOLACT algorithm?

denashamss commented 3 years ago

Hi @dbolya , thank you very much for the code you provided, actually I am using yolact for my project and I need to calculate precision and recall, is it possible for you to explain more on how can I print the precision and recall? thanks

MiguelAngeloMartinsRibeiro commented 3 years ago

@sainatarajan Hi, were you able to obtain the IOU? Thanks

MiguelAngeloMartinsRibeiro commented 3 years ago

@denashamss were you able to calculate IoU?

denashamss commented 3 years ago

@MiguelAngeloMartinsRibeiro no unfortunately still trying to find. Still have not received any update from @dbolya or others .

sainatarajan commented 3 years ago

@denashamss @MiguelAngeloMartinsRibeiro Sorry for the late response. Yes, I was able to get the IoU. Refer to my answer here. Once you save the masks, it's pretty easy to calculate the IoU. The code to calculate IoU can be taken from here.

unrivalle commented 2 years ago

Hi @dbolya I want to get the confusion matrix from eval.py .Can you please share some code to get it .

bhuvanofc commented 2 years ago

@sainatarajan Could you please share your code for Confusion matrix calculation

MiguelAngeloMartinsRibeiro commented 2 years ago

@sainatarajan I just saw your reply now, I'm having errors on your code. I'm in urgent mode for this, I'm delivering my thesis in a few weeks and my way to compute IoU is rudimentar and not that precise

I added a comment on stack overflow and my email is miguelangelomartins@tecnico.ulisboa.pt so if you can send me an email it would be perfect

Thanks for all your help

ARRARIAKU2 commented 1 year ago

@denashamss @MiguelAngeloMartinsRibeiro Sorry for the late response. Yes, I was able to get the IoU. Refer to my answer here. Once you save the masks, it's pretty easy to calculate the IoU. The code to calculate IoU can be taken from here.

hi @sainatarajan, how do i modify the code in eval.py and combine with the code that you gave from that link?

i still dont know how to get the IoU or can you put the full code of eval.py that you already modify?