Indoor climbing is an exercise that offers both physical and mental challenges and has, over the years, gained a vast amount of popularity. However, the difficulty of a climbing route is purely subjective, and factors like the climber's skill level and the type of climbing hold used impact the difficulty. Climbers use climbing holds, which are specially designed grips, to ascend a climbing wall. They come in various shapes, sizes, and textures, and their placement significantly impacts the difficulty and technique required to complete a route.

This study investigates the effectiveness of two approaches for detecting climbing holds in images. The first approach utilizes computer-vision with edge- and blob-detection. This traditional technique identifies potential holds based on the image's edge features and enclosed areas (blobs). In comparison, the second approach uses deep learning with the YOLO algorithm: This advanced approach uses a convolutional neural network (CNN) trained on labeled images to detect and classify climbing holds.

Our study's primary focus is on measuring and comparing two performance metrics. Accuracy: how well does each method identify climbing holds in the images? Execution time: What is the runtime of each technique used to analyze an image? In addition, we aim to identify both strengths and weaknesses with each of the approaches through evaluating the mentioned performance factors. In conclusion, we believe the gathered information throughout our study will be valuable when deciding which of the techniques, in the context of climbing hold detection, is the most suitable.

The outcome of the study yielded mixed results, YOLO achieved significantly higher accuracy in overall hold detection compared to the edge-detection algorithm. However, the YOLO model had a considerably slower execution time in comparison to using edge-detection. In conclusion, increasing the amount of training epochs offered marginal improvements, therefore YOLO is less suitable for applications that are resource-heavy.