Geraldine-M Mallet

The EOL Profile Newsfeed contains comments left for its owner by other members, EOL Community invitations, and gathers updates associated with the items in the owner's EOL watch list.

Add a new comment

  • Profile picture of Yan Wong who took this action.

    Yan Wong commented on "EOL Discussion Group":

    @Katja Schulz: Thanks. I've just checked and the original images were rotated correctly, but the rotation was done by setting the EXIF rotation metadata (see e.g. http://www.daveperrett.com/articles/2012/07/28/exif-orientation-handling-is-a-ghetto/). This metadata seems to be stripped out by EoL, not by Flickr. You can check because the original image on Flickr contains the correct rotation information (download if from http://farm4.staticflickr.com/3757/8812291168_ce56ecf228_o.jpg and have a look). I can re-upload my images on Flickr so that they are rotated without using the EXIF rotation hints, but I guess other people might get bitten by this too, as many cameras seem to use this way of rotating pictures. Maybe it should be on the (long) to-do list to get the image import routines in EoL to save a fully rotated image (as specified in the metadata) as the local EoL copy?

    about 1 year ago

  • Profile picture of Katja Schulz who took this action.

    Katja Schulz commented on "EOL Discussion Group":

    @Yan Wong: The original images you uploaded were probably incorrectly rotated, and they got auto-rotated by Flickr. However, we download the originals from Flickr, so we get the incorrectly rotated ones. Unfortunately we don't have a solution for this problem right now, so we have to ask our contributors to upload correctly rotated images to Flickr.

    about 1 year ago

  • Profile picture of Yan Wong who took this action.

    Yan Wong commented on "EOL Discussion Group":

    I've just tried posting some old pictures of mine to the EoL Flickr group. On Flickr the images appear in the correct orientation, but on EoL the imported images are sometimes rotated incorrectly. See for instance http://www.flickr.com/photos/96340474@N07/8812291168/ and its imported version http://eol.org/data_objects/24774483 I guess this should be a relatively simple bug to iron out. My suspicion is that it to do with the way image rotation is stored in jpegs by Mac OS X preview, but it could be a flickr auto-rotation thing.

    about 1 year ago

  • Profile picture of Cyndy Parr who took this action.

    Cyndy Parr commented on "EOL Discussion Group":

    @Yan Wong: We reharvest from Wikimedia Commons about once a week, so the change should eventually propagate to us. If it doesn't, let us know.

    about 1 year ago

  • Profile picture of Yan Wong who took this action.

    Yan Wong commented on "EOL Discussion Group":

    Does the EoL wikipedia scraping code revise currently existing EoL content depending on changes to wikipedia? I've just found this photo: http://eol.org/data_objects/17272302 which is the result of an accidental inclusion of the wrong species into the Wikipedia image gallery for Tremarctos ornatus. I've removed it from the gallery on wikipedia, but I don't know if I need to flag it up as misidentified on EoL too, or if my changes to wikipedia will propagate down to EoL.

    about 1 year ago

  • Profile picture of George Sims who took this action.
    George Sims joined the community "EOL Discussion Group".

    over 1 year ago

  • Profile picture of Sam Owens who took this action.
    Sam Owens joined the community "EOL Discussion Group".

    over 1 year ago

  • Profile picture of Cyndy Parr who took this action.

    Cyndy Parr commented on "EOL Discussion Group":

    @Yan Wong: Okay, I'll contact you by email and we can pick this up there. Nice work!

    over 1 year ago

  • Profile picture of Yan Wong who took this action.

    Yan Wong commented on "EOL Discussion Group":

    @Cyndy Parr:

    I've updated http://yanwong.me/MapImageTest.pdf with 225 new (green) data points pick haphazardly from monocots, composites, and pinophyta. The few green dots among the cloud of magentas are (I think) line drawings of plants. You can't see in this plot, but the png dimension completely separates out these plant line drawings from the maps, so yes - it seems to work almost better for plants. Dunno about microscopy images, though.

    Cyndy, I'll sort out a data file for you, so you can have a look in 3D yourself, if you like. Do you have an email address or whatever to which I can send the file? It won't be that big.

    over 1 year ago • edited: over 1 year ago

  • Profile picture of Cyndy Parr who took this action.

    Cyndy Parr commented on "EOL Discussion Group":

    @Yan Wong: Very promising approach! If you did some plants would you get the same results? I'm almost convinced we could use this to do some batch type changes.

    over 1 year ago

  • Profile picture of Yan Wong who took this action.

    Yan Wong commented on "EOL Discussion Group":

    Just looking for good images that are used in WikiSpecies, but not on EoL. Is there any reason why, for example, the nice image at http://commons.wikimedia.org/wiki/Condylura_cristata isn't being picked up. As far as I can see it meets the criteria at http://eol.org/info/curate_wiki#commons

    over 1 year ago

  • Profile picture of Yan Wong who took this action.

    Yan Wong commented on "EOL Discussion Group":

    @Cyndy Parr:

    I've written a little blog post about my first look at the problem of identifying maps. The plot on that page shows that just using the compression metrics seems to work extraordinarily well, at least for my limited dataset. http://yanwong.me/?p=709

    p.s. not sure this "map" that I picked up should be in EoL anyway: http://eol.org/data_objects/5853197

    p.p.s. Cyndy: I've just done a plot with all the images in your "Retag as map type" collection added (in magenta) - with this larger dataset just using jpg vs gif compression seems to work best, although adding PNG compression data as well would probably help remove a few of the false positives. See the plot at http://yanwong.me/MapImageTest.pdf - the numbers on each point are dataobject IDs, so you can find the corresponding pictures on EoL

    over 1 year ago • edited: over 1 year ago

  • Profile picture of Yan Wong who took this action.

    Yan Wong commented on "EOL Discussion Group":

    @Cyndy Parr: I've just had a quick play with classifying map images on EoL - since I've downloaded 738 pd/cc-by images to play with, of which 25 are maps (one for each placental mammal, as it happens). A first-pass is to see how well they compress using different methods. If you plot the compression ratios achieved using jpg and png compression, a good whack of the maps cluster in an obvious group. Have a look at http://www.yanwong.me/placentalmapclassification.pdf (images that are maps are marked in red). The EoL image name is written in tiny letters on each point, if you want to check up on them, but I don't know if these are unique names. It might take a few minutes longer for me to figure out how to put the EoL object id there instead. I've got a quick perl script that will calculate this in-memory, without saving jpg and png files to disk - you could try it on your maps collection. I've also got a few ideas for improving the clustering, but it seems a reasonable start.

    over 1 year ago

  • Profile picture of Yan Wong who took this action.

    Yan Wong commented on "EOL Discussion Group":

    @Cyndy Parr: Ah - thanks.

    over 1 year ago

  • Profile picture of Cyndy Parr who took this action.

    Cyndy Parr commented on "EOL Discussion Group":

    @Yan Wong: Nope, go to http://www.iucnredlist.org/search and search for Apis mellifera. It hasn't been assessed yet. Assessment is an arduous process and looks like only about 4000 insects have been looked at. I looked at a few and they were lepidoptera and odonates. Only 165 hymenoptera are on the list.

    over 1 year ago • edited: over 1 year ago

  • Profile picture of Yan Wong who took this action.

    Yan Wong commented on "EOL Discussion Group":

    @Roderic Page: Thanks Rod (by the way, I'm an ex-student of yours). Still puzzled that there isn't IUCN data for the default page search on Apis mellifera - I assume IUCN knows about this one :)

    over 1 year ago

  • Profile picture of Cyndy Parr who took this action.

    Cyndy Parr commented on "EOL Discussion Group":

    @Yan Wong: For historical reasons, we have two paths for info from IUCN...one for the status field at the top of the overview, and the other for the content on the details tab. I think the latter has the most up-to-date data even for the status field and we should look into getting rid of the former collection to avoid confusion.

    over 1 year ago

  • Profile picture of Roderic Page who took this action.

    Roderic Page commented on "EOL Discussion Group":

    @Yan Wong: The vast majority of species on the planet are not covered by the IUCN (nor are they likely to be). The IUCN site suggests some 65,000 species have been assessed http://www.iucnredlist.org/about/summary-statistics (see Fig. 2). EOL has two IUCN lists (why?), the largest http://eol.org/collections/309 has 63097 taxa.

    over 1 year ago

  • Profile picture of Yan Wong who took this action.

    Yan Wong commented on "EOL Discussion Group":

    @Cyndy Parr: Thanks a lot. I've joined the API community now. Is there any reason when I go to http://eol.org/api/docs/pages/1.0 and run the default test search but with IUCN=true, I don't get any IUCN data? I thought that every species had an IUCN classification?

    over 1 year ago

  • Profile picture of Cyndy Parr who took this action.

    Cyndy Parr commented on "EOL Discussion Group":

    @Yan Wong: Vernacular names data often comes to EOL dirty (or we mangle it, not sure which yet). Any curator can fix it by going to the common names tab such as http://eol.org/pages/328450/names/common_names (I fixed Mus musculus. Though I don't know the best French name I picked the one that had the most sources). If you regularly get scientific names as vernaculars, you might want to design your program to go back and get another vernacular. There are often other options in each language. Also, I invite you to join our API community. http://eol.org/communities/121/newsfeed

    over 1 year ago