 Hi, my name is Marc Lejeunesse, and I'm an ecologist at the University of South Florida and I'm gonna talk about three things today. One, I'm gonna talk about tools available to help researchers extract data from plots Then I'm gonna talk about how I fit into that space and some of the things I developed throughout the years and then three I Want to spend a lot of time describing how we could all improve these tools By advocating for really specific Plotting practices There's a lot out there to help you out extract data from plots One of the heavyweights is web plot digitizer. It's been around for a very long time. It's well maintained It has a very nice clean interface that helps you get what you need done In terms of our there's also many tools available Meta digitize is a popular one I myself have developed some manual and semi automated tools in Metagear Which has been around since 2015 I don't remember anymore at this point I mean it was the first tool that I created for the package was these data extraction functionalities and But the Super bummer part of creating all those tools And something that I've learned Sadly throughout the years is that nobody uses Metagear that way Nobody uses any of the manual or semi automated data extraction tools and I almost feel like I created some vaporware Because that really was a lot of the effort I put into the package was developing those tools and now I kind of understand why and So this is what I'm going to spend some of my time talking about is One making it easier for semi automated tools to become more common in extracting data from plots and to Again, like I like I like to end this This brief talk to Describe how we could all improve The use of semi automated tools by advocating for certain plot types Now Metagear Will soon have a Nice gooey interface that allow for the semi automated tools to be quickly done I feel like that was the number one barrier beforehand is Essentially, you would have to write a function to optimize the object detection functionalities in Metagear that was just like a turnoff for most Applied practices, you know, all you need is a digital ruler to point and click and measure stuff But once in a while, you'll get a high density plot and To make that quicker and more repeatable, you know, I feel like Metagear could help you out with that and so here for example You've been able to do the semi automated data extraction stuff, you know for like six years But now there's an interface to help you out to help you streamline that Parameterization of the object detection models But this is a super nice plot and the frequency unfortunately of super nice plots is Low What we are presented with is A huge diversity of plots and practices out there I mean maybe with just cause Because you know scientists get very few opportunities to be creative and so Presenting nice-looking plots is one of these things these avenues in which we could try to and Have fun, but a consequence of that is you break a lot of rules fundamental rules in plotting things With the hopes of making it easier for a reader to interpret your study outcomes that you're reporting in your paper But that there's a there's a trade-off associated with that where you made a plot super nice for a human but consequently it's a difficult for an automated tool to make sense of That to extract the data locked into the plot efficiently and in a repeatable way And so here are my pet peeves of plotting Which if we could avoid any of these types of plots it would Streamline the entire process. I mean we could develop automated tools much more quicker If we would avoid all of these silly silly silly practices And so basically I'm going to end the talk with a giant laundry list of things that add noise to the efficiency of semi-automated tools to extract data from plots GG plot is A troublemaker in the space. I mean it makes beautiful plots, but Vanilla GG plot does not present axes Visually right you got to tell it to to put in the axes without the axes. It's actually hard to start and Identifying the coordinates of objects in a space because you don't have that nice nicely defined boundary Divorced axes another our problem Vanilla are plots will do this It'll break apart the exact axes for some reason to maybe make it easier for you to understand what's being plotted Yeah, why? Text in plots. That's noise. That's flat out noise easy great for us to understand what those points are but awful for a Function or an algorithm to try to identify objects because now it has to differentiate between text and the points and That involves some awkward modeling this stuff. I don't know what you call this stuff This is becoming more and more popular where you Use images instead of points. Yeah, you got to figure out the centroid of that to extract data No, thanks high-density plots. Wow. That's a trouble to you Multiple plots within an image is a problem in lane plots is a problem Multiple plots is a problem and finally multiple y-axis plots are another problem. All these are problems. I Hope that in your future research and how you report things you avoid these issues. I'm working on Pulling together a guide best practice guide That is centered Not so much for the human right we're gonna keep some of the nice human features for readability But to also make it open for machines to read plots and a lot of it involves avoiding this stuff So if you're interested in learning about these details Contact me we could figure something out. Anyway Thanks for your time