epochs

[jiyeooong]

(R2D dataset)
How many epochs did you train on RoadSeg-152?
The default maximum epoch value is 1000. However it is too big size.
Thanks.

[hlwang1124]

Hi, @jiyeooong . Sorry that I do not remember the specific epoch value. One sure thing is that the network can converge within 200 epochs.

[jiyeooong]

Thanks!!!
I trained R2D datasets on these options.(Res34)
( --dataroot datasets/R2D --dataset R2D --name R2D --use_sne --num_threads 8 --nepoch 15 )
I only changed kitti_dataset.py code to R2D_dataset.py.
And I tested to my trained model on test datasets.
My results: Acc 99.1 /Pre 98.3 / Recall 99.4/ F-Score 98.9 / IoU 97.9
Why are my results better than your performance comparisons?

[hlwang1124]

Sorry for the late response. Did you adopt the same data split as mentioned in #2 ?

[jiyeooong]

Sorry for the late response. Did you adopt the same data split as mentioned in #2 ?

Yes. I did.

[syc10-09]

@jiyeooong Hi
After training 200 epochs, I got similar high result using resnet34. Have you found the reason why got this result? Could you tell me? thank you very much!!!