Accuracy

[amna-ai]

When I tested the I3D model for 100 classes, I achieved the following accuracy:
top-k average per class acc: 0.670542635658915, 0.8923076923076927, 0.9346153846153845
0.67 for top 1 class
0.89 for top 5
0.93 for top 10

It is more than the ones mentioned in the title of pretrained weights?
Did I do something wrong? How can I get more than mentioned?

[XiongTLu]

font{ line-height: 1.6; } ul,ol{ padding-left: 20px; list-style-position: inside; } Based on the WLASL paper, it is recorded in Table 4 that the top-10 accuracy of the I3D model trained with WLASL100 is 89.92%, so I think your training result is also very good and reliable.About your higher accuracy, I think it's because your hardware is newer than the author's. ***@***.*** On 8/23/2022 ***@***.***> wrote： When I tested the I3D model for 100 classes, I achieved the following accuracy: top-k average per class acc: 0.670542635658915, 0.8923076923076927, 0.9346153846153845 0.67 for top 1 class 0.89 for top 5 0.93 for top 10 It is more than the ones mentioned in the title of pretrained weights? Did I do something wrong? How can I get more than mentioned? —Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[hshreeshail]

Yeah. I am also getting better than reported results, though not as good as yours:
Metric: Top-1 Top-5 Top-10
Reported: 65.89 84.11 89.92
Mine____: 67.07 84.58 90.25
Yours____: 67.05 89.23 93.46

However, I am getting worse that reported results for 2000 classes
Reported: 32.48 57.31 66.31
Mine____: 30.45 55.36 63.87

Is this change simply down to hardware difference? Is it significant?

[amna-ai]

I don't know if hardware is causing this result. Did you find any other reason?

[hshreeshail]

No. Not sure what the reason for the discrepancy is.