search

mudit-dhawan / GAME-ON

1 stars 2 forks source link

Lower F1 score #1

Open CoisiniStar opened 9 months ago

CoisiniStar commented 9 months ago

Hello, I ran through your project, but since I failed to open the relevant pictures in RGB format, I cleaned the data set entries corresponding to the unopened pictures (Twitter), but the F1 value in the experimental data only remained around 50, may I ask if you have cleaned the data set? Looking forward to your reply, thank you

xiaozhi1233 commented 6 months ago

Hello, I ran through your project, but since I failed to open the relevant pictures in RGB format, I cleaned the data set entries corresponding to the unopened pictures (Twitter), but the F1 value in the experimental data only remained around 50, may I ask if you have cleaned the data set? Looking forward to your reply, thank you

你好，想问下imageGraph.py这个文件里的： image_t = transform(image).to(device) image_t = image_t.unsqueeze(0) # add a batch dimension image_t = image_t[:,:3, :, :] 这里的transform是定义的函数还是类？为什么没有其具体代码，运行会报错，您运行的时候报错了吗?

mudit-dhawan commented 6 months ago

Please use the same hyperparameters mentioned in the paper, it required a lot of tuning to get the best performance.

CoisiniStar commented 2 months ago

Thank you very much for your help, I have achieved some research based on your work and will follow up by sending that work to you privately, thank you again for your help.

CoisiniStar commented 2 months ago

Hello, I ran through your project, but since I failed to open the relevant pictures in RGB format, I cleaned the data set entries corresponding to the unopened pictures (Twitter), but the F1 value in the experimental data only remained around 50, may I ask if you have cleaned the data set? Looking forward to your reply, thank you

你好，想问下imageGraph.py这个文件里的： image_t = transform(image).to(device) image_t = image_t.unsqueeze(0) # add a batch dimension image_t = image_t[:,:3, :, :] 这里的transform是定义的函数还是类？为什么没有其具体代码，运行会报错，您运行的时候报错了吗?

你好，源代码确实是有问题的，直接将transform改为transforms即可，因为作者做了别名：import torchvision.transforms as transforms。希望对您有帮助！

VividBinGo commented 3 weeks ago

Thank you very much for your help, I have achieved some research based on your work and will follow up by sending that work to you privately, thank you again for your help.

您好，我如何对从 https://github.com/yaqingwang/EANN-KDD18 下载的微博数据集进一步处理，使程序能够成功运行imageGraphs.py或textGraphs.py？