no mAP score

[aliyyahna]

I tried training using my own custom dataset, everything seems fine until I check the record.csv file. Why are the mAP values zero? I followed everything from the program, am I missing something? Thanks!

[AlOo2002]

may i\i ask if you changed something in the code? I'm trying to train my own custom dataset too but the total loss result going upsait down

[aliyyahna]

may i\i ask if you changed something in the code? I'm trying to train my own custom dataset too but the total loss result going upsait down

I didn't change anything! I simply follow the whole code. I'm not really sure about the solution to your problem, maybe check the program again from the start?

[AlOo2002]

thanx

[adavradou]

Hello @aliyyahna! Did you find out why the mAP was zero? I face the same problem.

[YCAyca]

Because the code simply writes 0 for mAP calculation and go... It's not something calculated. So it's not important but it would be nice how to calculate mAP and write the real value...

new_row = {'mean_overlapping_bboxes':round(mean_overlapping_bboxes, 3), 'class_acc':round(class_acc, 3), 'loss_rpn_cls':round(loss_rpn_cls, 3), 'loss_rpn_regr':round(loss_rpn_regr, 3), 'loss_class_cls':round(loss_class_cls, 3), 'loss_class_regr':round(loss_class_regr, 3), 'curr_loss':round(curr_loss, 3), 'elapsed_time':round(elapsed_time, 3), 'mAP': 0}