12

Bib Recognition for Sports Photographers (Read 1621 times)

LedLincoln


not bad for mile 25

    Once the technology is in place, there are a lot of great options for it, just like you mentioned. There's even the possibility of having a photobooth at a race where people search for their bib# right after the race and print the photos an hour after they've finished. 

     

    This seems to me like it would be very popular.

      Thanks Blaine and Larry.

       

      It helps to hear that I'm not insane, and others see a potential here too. I have a number of additional photog's to chat w/about it so we'll see how it progresses. So far so good, so thank you!

      Creator of RunForth@robraux | shodless.com

      herzog


        Rob, this seems like a great application.  I would like to receive more information, as I would be a user and may also be interested in helping you sell this service in Brazil.   I just filled the information on your page, please, keep me updated. I will send you an email as well.

         

        Tks,

          Thank you hezrog, if you'd like to help out, a great first step would be providing valuable feedback on how you work now:  http://www.surveymonkey.com/s/FQKJX9H

          Creator of RunForth@robraux | shodless.com


          Patzer

            You can improve accuracy by fingerprinting the colors of the person wearing the bib. People generally stay the same colors while running a race (but not always). If you take it into account (e.g. Bottom=80% red, top=50/50 white/black) then you can increase confidence of candidate matches.

            Current goal: an Olympic distance triathlon. Did two of them! New goal: a half-ironman?? Did that too. New goal: I'll have to get back to you.


            Patzer

              (Even just noting the predominant color directly around the bib would help.)

              Current goal: an Olympic distance triathlon. Did two of them! New goal: a half-ironman?? Did that too. New goal: I'll have to get back to you.


              #2867

                I'm at the RRCA National Convention and one of the breakout sessions I attended was on new revenue sources for clubs and a lot of the presentation was about race photography.  Some of the questions came about for how to match photos to email addresses and the predominant response was, "We don't do that."  So, I looked up your domain on here and mentioned it to folks. Hopefully a few sign up for your updates.

                Run to Win
                25 Marathons, 17 Ultras, 16 States (Full List)

                  I did, thanks for the reference Blaine. I appreciate it. I've got about 4 more sign ups from interested folks!

                  Creator of RunForth@robraux | shodless.com


                  325th place or bust!

                    MT, photo-red type systems use an actual human to verify license plate matches before anything is sent out.  I can't speak for every state, but I know the two I looked into both required that human check.

                     

                    True, but photo-recognition systems have advanced to the point where they can be used in police cars to scan all license plates around them as they drive to identify ones that are on the wanted list, or have not paid registration, etc.

                    http://en.wikipedia.org/wiki/Automatic_number_plate_recognition

                    At the end they mention that systems tend to be 90 to 94% successful.

                    So it does look like there has been a lot of automatic number systems out there for quite a while, they just don't seem to be used for races yet.

                    PR: 5K 22:41, 10K 51:05, HM 1:59, Sprint Tri: done!

                      True, but photo-recognition systems have advanced to the point where they can be used in police cars to scan all license plates around them as they drive to identify ones that are on the wanted list, or have not paid registration, etc.

                      http://en.wikipedia.org/wiki/Automatic_number_plate_recognition

                      At the end they mention that systems tend to be 90 to 94% successful.

                      So it does look like there has been a lot of automatic number systems out there for quite a while, they just don't seem to be used for races yet.

                       

                      While quite true, they have a convenient set of restrictions available to them, including:

                      • Localization (standard length (and usually font)) of expected plates in the location.
                      • Standard colorization in the location
                      • The plate is almost always straight, or at a known angle.

                      On another downside, the systems are incredibly expensive. I'm not saying it can't work, just that I've never heard of anyone who has found it worthwhile. In fact I've heard the exact opposite.

                       

                      A specialized product for sports photographers can be optimized for the target market, just like license plate recognition has been. Each license plate photograph (or video often) is a unique moment. The chance of the car passing again is limited. A set of race photographs tell the race story. The same person will most likely appear 3-10 times.

                       

                      The ability to utilize this entire set of data, coupled with:

                       

                      • Shirt/pants color near the bib.
                      • Facial recognition.
                      • Expected bib numbers.

                      Allows the application to pull out and tag the photos in a manner and higher confidence level than would have been possible with just the photo and no context.

                       

                      At least... that's the theory I'm working off of :-)

                      Creator of RunForth@robraux | shodless.com

                      tagily.io


                        Hi! We have just released tagily.io, a cloud-based service for automatic bib number recognition in sport events.

                         

                        It's super easy to use and you'll have the results in a few minutes after you start uploading the pictures. We also provide a convenient upload tool. Depending on your internet connection speed, we can process up to 10.000 images per hour.

                         

                        As the results come back, you can choose to refine them with our annotation tool (free to use) or make them available as they are. You can use it on all types of bibs, no additional code placement is required (e.g. QR or Aruco code), because we read the number directly. We don't even have camera requirements: as long as the bib number is good enough for a human to read it, our artificial intelligence system can also read it.

                         

                        As of today, our biggest event has been the Marcialonga 2018, where 100.000 images were processed in one night. Yet the software smoothly works for marathons and bike races as well. You can learn more about it here: http://www.deepvisionconsulting.com/automatic-bib-recognition-kick-off-at-marcialonga-2018 or just go to the website to ask for a free trial: https://tagily.io.

                        12