Skip to content

Commit

Permalink
- Markdown fixes.
Browse files Browse the repository at this point in the history
  • Loading branch information
mikemaccana committed Oct 20, 2010
1 parent 36ce861 commit cdfc3fe
Showing 1 changed file with 11 additions and 9 deletions.
20 changes: 11 additions & 9 deletions README.markdown
Original file line number Diff line number Diff line change
Expand Up @@ -10,20 +10,13 @@ It provides a simple True/False response to whether an image is adult related. I
- User generated content sites
- Content-filtering proxy scanners

### General thoughts on PIFilter

In my own experience as a customer PIFilter's service is like health product that works best with diet and exercise. It is useful as a basis to put specific posts ahead in a moderation queue, where the suspect posts are checked by a human. But it's not sufficient on its own to replace human moderations completely.

With my own user generated content tests, accuracy was generally good with a tendency towards False positives rather than False negatives.

Of course, your milage may vary as you'll be working with a different set of input, so best to see for yourself.

### Filtering images

- Get a [pifilter account and customer ID](http://www.pifilter.com/).
- Then [download pyfilter](http://github.com/mikemaccana/pyfilter/tarball/master).
- Use **pip** or **easy_install** to fetch the **urllib2** and **poster** modules.
- Then run:

Then run:

from pyfilter import checkimage

Expand All @@ -37,6 +30,15 @@ By default pyfilter runs in 'aggressive' mode, where 'maybe' responses are count

which will count 'maybe' responses as False.


### General thoughts on PIFilter

In my own experience as a customer PIFilter's service is like health product that works best with diet and exercise. It is useful as a basis to put specific posts ahead in a moderation queue, where the suspect posts are checked by a human. But it's not sufficient on its own to replace human moderations completely.

With my own user generated content tests, accuracy was generally good with a tendency towards False positives rather than False negatives.

Of course, your milage may vary as you'll be working with a different set of input, so best to see for yourself.

### We love forks, changes and pull requests!

- For this project on github
Expand Down

0 comments on commit cdfc3fe

Please sign in to comment.