search

PolynomialQian / RD3D

RD3D: RGB-D Salient Object Detection via 3D Convolutional Neural Networks
MIT License
49 stars 8 forks source link

Can you provide the RD3D results of ReDWeb-S? #10

Open fereenwong opened 2 years ago

fereenwong commented 2 years ago

Hi, thanks for sharing. Can you provide the RD3D results of ReDWeb-S?

PolynomialQian commented 2 years ago

ok, the RD3D results of ReDWeb-S have been updated～

fereenwong commented 2 years ago

Thanks a lot. I download and evaluate the updated RD3D results of ReDWeb-S, it results in a performance that differs from the one listed in Table. (RD3D: ReDWeb-S 0.700). Are the updated results consistent with the one used to report the performance?

PolynomialQian commented 2 years ago

Yes, since we did not retrain the ReDWeb-S dataset and only tested it on this dataset with the previous model, this result is to be expected.

At 2021-11-23 14:57:52, "fereenwong" @.***> wrote:

Thanks a lot. I download and evaluate the updated RD3D results of ReDWeb-S, it results in a performance that differs from the one listed in Table. (RD3D: ReDWeb-S 0.700). Are the updated results consistent with the one used to report the performance?

— You are receiving this because you commented. Reply to this email directly, view it on GitHub, or unsubscribe. Triage notifications on the go with GitHub Mobile for iOS or Android.

fereenwong commented 2 years ago

Sorry for the misunderstanding. What I want is the results of the previous models. As I want to add this as a comparison method in my paper, which keeps the same setting as the previous models (not trained with the trainset of ReDWeb-S). I tried to get it with the provided pretrained model, maybe I did something wrong, I can not exactly reproduce the performance as reported. Could you please send me a copy of the ReDWeb-S results with previous models? Email: fereenwong@gmail.com. This will help me a lot. Thanks again.