YuLengChuanJiang

@FredlinT @yangyang12315 Hello, did you know the reason why the pre-trained model (rgb_0.946) tests on the ucf-101 split1 but its accuracy is around 0.91? Can you give me an answer?...

that's right, but you also need replacing os.path.exists(os.path.join('UCF-101', classname, filename) with os.rename(os.path.join('UCF-101', classname, filename), dest) i think the author may make wrong uploading the .py file.

@ilkarman hello, have you reach the accuarcy that descirbed in the repo on the UCF101 using i3d, I can only reach 0.89 for RGB on split 1.

@innerlee @ADstudyer I also encountered a similar problem, which occurs when the frame count is low. Can you share your solution? 

您好！您的邮件我已收到，我会尽快查看并回复 ，请不用担心。 祝您万事如意！                                                                         冷川江

@joaoluiscarreira Hello, Joao. I want to know how much storage space is required to extract RGB and flow frames from all Kinetics400 videos. Best

@Tyler-D @bingnoi Hi, I'm working on action recognition. I wonder to know that have reproduced the I3D model on the UCF101? And how about the accuracy?