[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [smila-user] SMILA vs Human Computing
- From: Ilio Catallo <ilio.catallo@xxxxxxxxx>
- Date: Thu, 22 Dec 2011 17:31:34 +0100
- Delivered-to: firstname.lastname@example.org
- Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=from:mime-version:content-type:subject:date:in-reply-to:to :references:message-id:x-mailer; bh=ykS2UC9x/4zoEakZj3SX3YcXl8tw6++rU8J/0GnPjWY=; b=AcWC3uRzYH/2AeD9fkenXhG5fQxRq8KqILYMfGfnsSiWtt+8O/QNRZeeDxV69FUezC 1UZaWB08gaMxX5XTw2CqS8AU9/arbm7opTQt0IDmbFSf/MbsE39WEZ35r7lMjntnKvkB yik5T+SryHEWAHiXsf20UA1n3ikI6ZiFo+vis=
We thought how to circumvent the problem. To evaluate the feasibility of those workarounds we would like to know:
- Is it possible to set the pipeline timeout equal to infinity? If so, which is the file that has to be modified?
- Is it possible to use within SMILA a BPEL pipeline such as the one in the attached figure? In this case we would be able to wait for an event sent by the crowdsourcing webapp and let the pipelet P2 modify the content indicated by the users of the crowd. We would like to know the technicalities that stay behind this approach.
Thanks in advance,
Description: Binary data
Il giorno 21/dic/2011, alle ore 09:22, Jürgen Schumacher ha scritto:
SMILA has not been developed with such use cases in mind, so I do not suppose that anyone has already created such a pipelet. Also I don’t think that a pipelet would be the right place to do something like this. For example, there are limits on the execution time of pipelines so you just cannot wait arbitraty long for some user to pick up the request. Maybe it would be possible to do it as a job worker, but I think that could be quite complicated, too. And the jobmanager is not designed for something like this, too. I suppose you would have to implement the actual user interaction outside of SMILA: a pipelet could add the logos to check to some database (or folder), where a web application can pick them up and present them to a user, the user can make the decision and the true logos are added back to SMILA by invoking another pipeline or adding them to another job for asynchronous processing. Both could be easily done using the SMILA ReST API.
In the last few days at Politecnico di Milano we were trying to develop an entire video analysis pipeline.
I'll briefly summarize the pipelets:
- A first pipelet downloads some logos from Google Images (es. Coca Cola logos) and stores them into a folder
- A second pipelet compares each found logo with each frame in a video collection, trying to identify all the frames that contain one of the logos
We want now to integrate a third pipelet that stays between the first and the second one above mentioned.
This third pipelet takes all the downloaded logos from the folder and displays them onto a web page. At this point, a user in the crowd recognizes whether the displayed logos are true logos or not. The true logos are selected (e.g. with a checkbox). Once the selection is made and the request is submitted, the logos that weren't selected are removed from the folder and the process continues with the execution of the third pipelet.
I wanted to ask you whether there's an implemented example for something that resembles this pipelet, or whether you can help us in the implementation in case.
smila-user mailing listsmila-user@xxxxxxxxxxxhttps://dev.eclipse.org/mailman/listinfo/smila-user