search

xuehy / Cross-View-Gait-Deep-CNNs

A Comprehensive Study on Cross-View Gait Based Human Idendification with Deep CNNs
45 stars 12 forks source link

The test accuracy and time cost #6

Closed wwyfzu502 closed 5 years ago

wwyfzu502 commented 5 years ago

Thank you very much for sharing! I still have some doubts. The first one is about the result table you offered in read me. It seems that there is a gap between your result and the paper's(showed below) image Is it the best result you got? How many iterations did it take？

And in your answer for another issue,i notice that you said you spend around 10 hours in GTX 1080 Ti(the same to mine). Did you mean you take 2000K(stop num in your code) iterations in around 10 hours? It seems that 6 days will be spend in my device.

wwyfzu502 commented 5 years ago

I made a mistake, 3 days will be spend in my device. I would be very grateful if you could answer my doubts.

xuehy commented 5 years ago

I did not try to replicate the results in the paper. To improve the results, you may need to tune the parameters and wait for the model to converge. Besides, my evaluation method may be different from the original paper's evaluation approach. I think you may check if the evaluations are the same.

wwyfzu502 commented 5 years ago

All right, I get it. Thank you very much