[FFmpeg-user] Command being killed in perceptual hashing / ffmpeg in general

Carl Eugen Hoyos ceffmpeg at gmail.com
Fri Oct 20 01:46:01 EEST 2017

2017-10-19 17:22 GMT+02:00 Aoife Fitzmaurice <afitzmaurice at irishfilm.ie>:
>> Does it make sense to use the signature filter on two files with very different length?
> The purpose for which I am using the script is to compare two files for similar content, the shorter of which I believe is a partial dupe of the longer.
> Andrew Weaver has written an interesting blog on the subject: https://ndsr.americanarchive.org/2017/04/20/adventures-in-perceptual-hashing/

(Given that kind people provide current binaries, the link to the
extremely complicated build page does not seem useful.)

> Up until now I've been generating and comparing checksums, but the perceptual hashing script would definitely enhance our archival workflow.

Allow me to repeat my question (the blog is a little long and as said
above its links do not make it more credible):
Does it - in general - make sense to compare two files of (very)
different length with perceptual hashing?
You definitely have found a bug but I still find it interesting if the
issue you reported happens for a real-world
use-case or only for pathological cases.

In general, OOM can happen with FFmpeg and real-world use cases so
while we try to avoid huge
memory allocations they may not be avoidable (this is not true in your
case though afaict).

Carl Eugen

More information about the ffmpeg-user mailing list