tianyu0207 / RTFM

Official code for 'Weakly-supervised Video Anomaly Detection with Robust Temporal Feature Magnitude Learning' [ICCV 2021]
326 stars 78 forks source link

feature dimension mismatch issue #78

Open coranholmes opened 2 years ago

coranholmes commented 2 years ago

您好,我加载了您提供的i3d features,但是发觉和原本的视频有些对不上。例如 ./save/Crime/UCF_Train_ten_crop_i3d/Normal_Videos586_x264_i3d.npy 原视频有1588帧,1588/16=99.25,上取整之后是100,但是您提供的视频维度是120(其他维度略),请问是怎么回事呢? 我统计了训练集里所有mismatch的视频如下(第一个视频是您的维度,第二个是我计算的维度):

./save/Crime/UCF_Train_ten_crop_i3d/Normal_Videos586_x264_i3d.npy 120 100
./save/Crime/UCF_Train_ten_crop_i3d/Normal_Videos600_x264_i3d.npy 158 132
./save/Crime/UCF_Train_ten_crop_i3d/Normal_Videos635_x264_i3d.npy 85 71
./save/Crime/UCF_Train_ten_crop_i3d/Normal_Videos644_x264_i3d.npy 91 76
./save/Crime/UCF_Train_ten_crop_i3d/Normal_Videos645_x264_i3d.npy 67 56
./save/Crime/UCF_Train_ten_crop_i3d/Normal_Videos648_x264_i3d.npy 67 56
./save/Crime/UCF_Train_ten_crop_i3d/Normal_Videos649_x264_i3d.npy 68 56
./save/Crime/UCF_Train_ten_crop_i3d/Normal_Videos650_x264_i3d.npy 82 69
./save/Crime/UCF_Train_ten_crop_i3d/Normal_Videos651_x264_i3d.npy 87 73
./save/Crime/UCF_Train_ten_crop_i3d/Normal_Videos653_x264_i3d.npy 66 55

测试集里所有mismatch的视频如下:

./save/Crime/UCF_Test_ten_crop_i3d/Normal_Videos_901_x264_i3d.npy 74 73
./save/Crime/UCF_Test_ten_crop_i3d/Normal_Videos_913_x264_i3d.npy 39 38

除了这几个视频,其他的维度都能对的上请问是怎么回事呢(:з」∠)

The same issue was mentioned in #22 #36