/members-area/personalblogs/b/blog/posts/applying-ai-to-climbing-deep-learning-meets-the-odd-human-datasetIn this blog series, I attempt to bring two of my passions together … AI and rock climbing. In my last blog, I talked about the 2020 Tokyo Olympics that are expected to take place in 2021, and which will feature rock climbing as a discipli...en-USTelligent Community 12https://community.element14.com/members-area/personalblogs/b/blog/posts/applying-ai-to-climbing-deep-learning-meets-the-odd-human-datasetFri, 16 Apr 2021 17:44:12 GMT93d5dcb4-84c2-446f-b2cb-99731719e767:c0e674f8-5914-453c-98c8-54c722b1486aralphjy0<p>Just some random thoughts...&nbsp; </p><p>&nbsp;</p><p>I&#39;ve mentioned it before, but it seems like too hard a problem (too many variables) to solve without a huge amount of data.&nbsp; What if you could augment the data with a few sensors that would attach to equipment that is requisite to all climbers - left shoe, right shoe, rosin bag.&nbsp; Haven&#39;t thought of how to implement it, but then detecting orientation would be easier.&nbsp; Not sure how to locate hands (maybe wristbands?)... and obviously this would only work with future climbs.</p><p>&nbsp;</p><p>For video only, what if for a specific climb you could train the model to identify each climber.&nbsp; I&#39;ve seen cases where data is augmented by taking a specific image and rotating it.&nbsp; Then your training data set would have each person with varying orientations.&nbsp; And if you had different profiles that would be even better.&nbsp; Not a generalized case, but seems like it might work.</p><p>&nbsp;</p><p>Hope that you can find an existing dataset.&nbsp; Maybe you can start one <span>[View:/resized-image/__size/16x16/__key/commentfiles/f7d226abd59f475c9d224a79e3f0ec07-c0e674f8-5914-453c-98c8-54c722b1486a/contentimage_5F00_1.png:16:16]</span>.</p><img src="https://community.element14.com/aggbug?PostID=11167&AppID=293&AppType=Weblog&ContentType=0" width="1" height="1">