Snapshot Serengeti Talk

Will you be able to discern good classifyers from not so good?

  • tirralirra by tirralirra

    Will you be able to discern good classifyers from not so good? Can you eliminate any pranksters? (Not that I expect there would be many.)

    Posted

  • kosmala by kosmala scientist

    Short answer: yes, we'll be able to detect people willfully trying to make wrong guesses.

    Posted

  • Dov by Dov

    it would be fun to know how well we are doing - how well we are identifying. will this ever be a feature?

    it could be very simple - like what percent chance do i have of correctly identifying at least one of the animals in the photo, where "correct" is arbitrarily defined as what the majority of people chose for that pic.

    Posted

  • Dov by Dov

    also it would be fun to know how many pics we have identified, that would be simple enough to add i think. then we could compete with friends!

    Posted

  • libervurto by libervurto in response to Dov's comment.

    There is a record of your contributions on the main zooniverse page.

    Posted

  • Bennett_Brown by Bennett_Brown

    I like the idea of knowing how many I've contributed and to know my accuracy, @Dov. @libervurto, the zooniverse profile only says your own contributions. A ranking or leader board leverages people's competitive side for the benefit of crowd sourcing. Even non-competitive people respond to rewards, whether token silly rewards like fun noises from a video game or a splashy animation or token silly rewards like position on a leader board.

    Once a consensus identification is calculated or established by an expert, I'd love to know what percent of my answers were deemed correct.

    It would be nice to be able to filter my "recents" by animal.

    Posted

  • tillydad by tillydad moderator

    This is one that the scientists may like to answer 😃

    Posted

  • aliburchard by aliburchard scientist, translator

    @Bennett Brown, we've found that adding competitive elements tends to have a net neutral or negative effect, as folks who compete are those who were already motivated, and folks who would otherwise participate become discouraged because they feel they have no chance of "doing enough." But it's interesting to hear the demand for measures of individual accuracy - something we'll keep in mind!

    Posted

  • tillydad by tillydad moderator in response to aliburchard's comment.

    Thanks for the reply @alibuchard . I agree that it may demotivate people if there were to be a performance review and it is more important to have a wide range of people classifying 😃

    Posted

  • dodone by dodone

    My reward is the virtual safari I get every time I log on. I understand more about the Serengeti and its wildlife from this site than from all the drums and music documentaries.
    The incentive to be reasonably accurate with the ID is twofold: I'd hate to be blocked for 'poor performance and there is a scientific purpose behind the pictures that can be enhanced with our humble input.

    Posted

  • davidbygott by davidbygott moderator in response to dodone's comment.

    Thanks @dodone - we appreciate what you give to the project, and are glad that you find it so rewarding 😃

    Posted

  • rlb66 by rlb66

    Based on the incorrect IDs shown in the "Computer Vision Serengeti" study, I would question any personal accuracy stats. If I am right then the moderators, who are almost always correct, would have a much lower accuracy then would be expected. I can not believe, based on the mentioned study, that the average person is close to matching the experts.

    Posted

  • maricksu by maricksu moderator

    Every answer is valuable, needed and highly appreciated in this project. I believe we all here have tried our best and we all should be very proud of ourselves and happy for our great contribution.

    I add here some links (from the Blog) where info, how this project works and about the great study results that have been achieved thanks to our help.

    https://blog.snapshotserengeti.org/2012/12/14/we-need-an-i-dont-know-button/

    https://blog.snapshotserengeti.org/2015/06/09/snapshot-serengetis-first-scientific-publication-today/

    https://blog.snapshotserengeti.org/2016/05/10/hot-off-the-presses-get-your-good-data-right-here/

    https://blog.snapshotserengeti.org/2016/11/17/sciencing/

    About the learning identifying these images, the more we classify, the better we get. We can learn something through every image. Learning in everything takes some time, includes mistakes and that is completely ok. Many ways for us to improve skills, following on Talk what others have asked, reading previous comments in images, comparing images, looking at collections of other volunteers, on Search-page is possible to search for more images of species (nice way to see animals e.g. in various postures), reading discussion-threads (site is full of interesting, useful info), asking questions etc. Seasons may be very different and no matter how well we are able to identify animals in general, it takes a bit time to get familiar with e.g. backgrounds in images and specialities in each season. Also possible quality issues in images, different camera-types used or heights they are set, add the challenges. I take these challenges as the joy of learning.

    I have found this project very rewarding when having this chance to learn more, enjoy wonderful views and animal findings, and work together with all of you to contribute this great project and the future of this unique nature. 😃

    Posted

  • rcooley001 by rcooley001

    I would like to say thank you to the moderators, for there quick responses, it really helps to have species confirmation on some of these that are difficult to see. I also appreciate the kind way in which you correct inaccuracies, I have worked on other projects where folks have not been so kind, and it takes a lot of the joy out classifying. Half the fun is that when through trial and error and the acquisition of knowledge that results from examining thousands of images, one is able to instantly identify and differentiate a topi from an eland etc. Thank you for the opportunity to be part of this great project!!

    Posted

  • maricksu by maricksu moderator in response to rcooley001's comment.

    Thank you @rcooley001 for your very kind message, makes really happy and it is very much appreciated. Happy to help whenever can and possible. So glad that you feel this way and enjoy learning and doing your great work in this project, which can give us all so much in many ways. Very happy for this opportunity to be part of this project with all you wonderful people across the globe 😃

    Posted

  • davidbygott by davidbygott moderator in response to maricksu's comment.

    Ditto from me too! Maricksu you are wonderfully kind and enthusiastic, a great asset to this project 😃

    Posted

  • maricksu by maricksu moderator in response to davidbygott's comment.

    Thank you so much David, so great to be part of this together with all of you! 😃

    Posted

  • rcooley001 by rcooley001

    You are truly welcome.

    Posted