search

OrdinaryQin / DST-HCN

4 stars 0 forks source link

is there something wrong with my experiment #1

Open OskarIkaAdi opened 1 year ago

OskarIkaAdi commented 1 year ago

Hello, I'm trying to do an experiment using your code. but I got a result for NTU RGB+D X-Sub of 87.29. I am using NVIDIA GeForce RTX 3090 graphics card and batch size: 22, number of epochs: 90

Unlike your result of 92.3. Is there something wrong with my experiment. Can you enlighten me?

Thank you

OrdinaryQin commented 1 year ago

Hello, thank you very much for your interest. First of all, regarding your precision fusion question, I have it on file, and since I'm currently out of town, I'm waiting to reorganize the information when I get back. Secondly, regarding your experimental configuration, the batch we used should be 64, and we used multiple M40s for the experiments, and in the additional experiments I found that different experimental resources would lead to a slight error. Also you can make sure about other issues such as consistent processing of the dataset. What can be guaranteed is that our results are 100% trustworthy. Thanks again for your attention.