Closed DonaldTsang closed 4 years ago
Hi @DonaldTsang -- We've just published v0.3.0 which includes more support for video hashing.
With respect to timestamps, the SimpleSceneDetection
hasher wrapper now includes a compute_batches
method which allows access to the timestamps for hashes.
Given that this issue is somewhat general and that video support has been expanded, what do you think about closing this issue in favor of separate, more granular issue(s)? Those issues could focus on specific hashing algorithms you find promising and for which you might want to consider contributing an implementation via a PR.
Thanks for the reply. It would be good from now on to dig through different papers to see if it is possible to have advanced video deduplication algorithms, as I am sure there are a lot of potential i that field. Contributions would be a great way to participate as well.
There have been proprietary apps like https://www.video-comparer.com/ The idea is: Given videos A and B, is it possible to match which parts of A aligns with which parts of B? Also, it might be good to provide timestamps so that it can be input into MPV for close comparisons.
For other references: https://github.com/TorresAndy/4Hash/issues/2
There are forks of https://github.com/jonasob/blockhash/network that has such feature. pHash already includes video hashing https://github.com/aetilius/pHash There has been researches like https://github.com/rednoah/VASH and https://github.com/muesli/xray and https://github.com/kristiankoskimaki/vidupe and https://github.com/sschnug/pyVideoHash https://github.com/sschnug/pyVideoPartialCopyDetector and https://github.com/lsfxz/phidup and https://github.com/mediamicroservices/mm as well.