matjojo/tagrank

TagRank is a Hydrus API script that uses MS TrueSkill to figure out which tags you like best.

tagrank's Issues

ratings.json is not generating

The ratings.json does not generate when closing the software and manually creating an empty file with "ratings.json" as the name crashes the program.

Quality issues/open externally

There are some problems with the quality of the displayed pictures at times, most notable on large pictures with lots of small text.

I think adding a feature to click on or otherwise select an image could open it in an external program or in hydrus, allowing access to photo viewing controls without having to implement things like zooming and panning yourself.

Passing choices

How does the system currently handle non-decisions? How does that impact the ratings?

Does the system prioritise getting decisions for images that haven't been shown as often, or is it fully random?

Picture ratings

The comparison system is absolutely amazing, in terms of controls and selections and usability.

I was wondering if it'd be possible to rate individual images using similar code and the same comparisons. The software already somewhat does this on accident when an image has a unique title tag. It'd just be a matter of controlling the value of a rating metric (probably inc/dec) in the database. Trying to rate thousands of images off of personal taste is very difficult without algorithmic help, and I feel like this software or something very similar could help.

Tagging

In regards to the tag ranking, results are thrown off by websites that don't import tags,
Perhaps the data needs to be adjusted to accommodate?

But on the other hand, a tool to assist in tagging or setting up tag relationships would be very beneficial.