Closed ZiruiYan closed 3 years ago
Hi there, one of the way extracting is using the ffmpeg.
ffmpeg: https://ffmpeg.org
ex) ffmpeg -i 'input_file' -r 30 'result_file' # 30 means the fps, you can check the video info by using the ffmpeg -i 'video_file'
Hi Thanks for your reply! I try ffmpeg -r 1 -i 'input_file' -r 1 'result_file', which will automatically use the video's fps. And found that the fps may be different for different videos. eg. 04_001.avi is 25 fps while others are 24 fps.
Use the code above I extracted 274515 frames, which is the same as the paper "A Revisit of Sparse Coding Based Anomaly Detection in Stacked RNN Framework". But the provided test set has 40791 frames, which is less than 42883(given in the paper). I wonder if you did additional processing on this data set? Or I can access the original video of the test set.
Use ffmpeg -r 1 -i 'input_file' -qmin 1 - qscale:v 1 -r 1 'result_file' This will make the image size the same as the test set provided
Hello, can you provide the download link of the processed shanghaitech dataset?
使用 ffmpeg -r 1 -i 'input_file' -qmin 1 - qscale:v 1 -r 1 'result_file' 这将使图像大小与提供的测试集相同 The original test set was originally just images, with only 40791 frames. Why can the test set in the article reach 42883 frames?
I found that all videos from ped1, Avenue and the test set of Shanghaitech of science and technology were extracted into frames. However, Shanghaitech's trainsets are in video format. How to extract it?