As personal fitness has risen in popularity, so too has the use of wearable fitness devices. There exist products for many sports, but few for Indoor Rock Climbing. Current products which provide post-climb reports utilize specialized hardware equipped with sensors. As such, they are limited in the types of information they can gather. We develop a novel tool which makes use of Computer Vision and Machine Learning to analyze video footage of the climber and produce post-climb reports that convey information which cannot be ascertained through sensor-only methods, such as the percentage of route completed and number of moves taken. To do so, we identify climb holds using Object Detection models such as YOLO, as well as the position of the climber using Keypoint Detection APIs. Our current approach produces post climb reports with an average RMSPE of 0.266 and we highlight specific areas of improvement to bolster the performance. Such a tool will provide a simple way for rock climbers to receive feedback on their climbs, with the use of a standard cell-phone camera
We plan to create the Rock Climbing Coach, which is a tool that will help rock climbers. Rock climbing has been growing in popularity in recent years, even making its first appearance in the 2020 Olympics. Our project will be a tool that will help both new and experienced climbers better understand their climbs. The future vision is to have a system set up in a rock climbing gym giving feedback to climbers after every climb. This would help beginners to get off the ground, and help experienced climbers reach the next level of improvement. Throughout Spring 2022 quarter we will be working on a proof of concept for the Rock Climbing Coach. We plan to create a program that’s given a video of someone rock climbing and outputs a "climb report" which provides analysis on the climb.
We will be creating a program that takes in a video of someone rock climbing and returns a climb report. The climb report will provide information on how many moves the climber made, time of the climb, percentage of climb completed, and efficiency of the climb. The project is broken up into two parts: processing of the video, and analysis of the climb. We will be dividing up the processing of the video into two parts: hold detection and pose estimation. Team 1 will be working on hold detection using a combination of computer vision and machine learning approaches. Team 2 will be working on pose detection using machine learning approaches and other algorithms. For the analysis of the climb report, we will also be dividing up the different aspects among the two teams. Team 1 will focus on the aspects related to the rock climbing holds, and team 2 will focus on the aspects related to the pose and analysis of positions. All of these aspects will be combined to create a final climb report.
Detection of all the holds on the wall to identify percentage of route completed and move validity.
Pose estimation of the climber to identify number of moves taken, distance center overall moved and time elapsed.