Whipper - Giving Climbers a Hand

Inspiration

Our team was formed as a group of friends from the UVA Rock Climbing team. Each of us shares a passion and love for the rock climbing sport, so it only made sense our hack was climbing-related. The climbing community is incredibly diverse and inclusive and it is truly a sport for climbers of all shapes and sizes. But! Is there a certain type of body that naturally makes you a better climber?

What it does

We decided to look at climbers' hands to see if there were certain features that make it easier to climb more difficult grades. We hope that eventually, our application will also be able to predict athletes climbing potential through a picture of their palms.

How we built it

Our hack consists of three main milestones. First, collecting data. There was a lack of image data on the internet for us to parse, so decided to collect our own. We stayed at the UVA Slaughter Recreation Gym to ask climbers for images of their hands as well as the climbing grade. In total, we collected 30+ data points. The second is image processing. To limit confounding variables we processed each image by resizing, deleting background, and turning the hand into greyscale. Third, neutral network. After processing each image, we ran a script that analyzed the shape of the hand to determine the grade of the climber.

Challenges we ran into

Our biggest challenge was dealing with an extremely small dataset. All the data we used for this hack was collected during the duration of the hackathon. Therefore, when parsing through our images in our neural networks we returned high overfitting values.

Accomplishments that we're proud of

This is our team's first hackathon! We are proud to have shown up and let alone submit a hack. We are super excited to reflect on how much we learned as well as all the people we have met in the process.

What we learned

While we learned plenty of technical skills such as VSCode, OpenCV, Tensorflow, Spyder, we would like to take away the fact that completing a project given such little time as possible for us. We learned that sometimes its okay to not know what we are doing and just keep going at it!

What's next for Whipper - Giving Climbers a Hand

The next step for Whipper will be to collect more data. We hope that we could turn this into an application that could be accessed by any member of the rock climbing community. It would be a place where rock climbers could input their own datasets. Over time the application would be able to become more and more accurate.

Built With

• opencv
• spyder
• tensorflow
• vscode