hadoop and video processing

[Oleg Ruchovets <or...@gmail.com>]

Hi ,

The internet and google gives me preatty poor result regarding VIDEO
PROSESSING / ANALYTICS,
so in case someone could point me on intersting links or best practices
that would be great.

My vision is to proces one file per one map taks and result pass to
reducer.
     Am I right  that in order to work 1 file per map I need to write
Custom InputFormat or there is another way to get one file per map?

Thanks in Advance.
Oleg.

Re: hadoop and video processing

[Bertrand Dechoux <de...@gmail.com>]

I am not an expert in distributed video processing / analytics. However,
the answer you are seeking will depend on the size of your videos and the
algorithms you want to use. Ideally, you would want to process part of the
videos in parallel using map tasks and then merge the results using the
reduce tasks. If it is not feasible you indeed should enforce one video to
be processed by the same map task. However, that does not necessary mean
that each map task should only process a single video (that is a stronger
constraint which would have a bigger negative impact on you overall
performance).

If you provided more information about the size of the videos you want to
process and the kind of processing you want to do, you might receive more
precise answers.

Regards

Bertrand

On Mon, Sep 24, 2012 at 5:25 PM, Oleg Ruchovets <or...@gmail.com>wrote:

> Hi ,
>
> The internet and google gives me preatty poor result regarding VIDEO
> PROSESSING / ANALYTICS,
> so in case someone could point me on intersting links or best practices
> that would be great.
>
> My vision is to proces one file per one map taks and result pass to
> reducer.
>      Am I right  that in order to work 1 file per map I need to write
> Custom InputFormat or there is another way to get one file per map?
>
> Thanks in Advance.
> Oleg.
>



-- 
Bertrand Dechoux