The inspiration for this project came from our desire to use machine learning to process images and produce a prediction. Our original idea was to feed the computer pictures of various insect bites in order to provide a quick diagnosis for individuals who may notice some suspicious redness or mark. Due to a lack of a proper data set and minimal time, we switched our focus to the problem of posture, specifically of the pelvis. This posture problem was something we all noticed and have thought about in our everyday lives. Pelvic tilt can lead to other future issues, including chronic lower back pain, hip pain, neck pain, and shoulder pain.
Our hack takes in a photo of a user in a side profile and, utilizing machine learning, we offer a potential diagnosis for whether the person in question has a posterior pelvic tilt, anterior pelvic tilt (the most common), or no tilt at all.
Having next to no experience, it was a steep learning curve to understand the code and technology we were trying to use and implement it in such a short period of time. We were all friends from high school, but we hadn’t worked together on a serious project before. As a result, it took some time to adjust to the resulting dynamic such as equal work distribution and meeting the deadlines we set for each other. Luckily, our camaraderie allowed us to reconcile our issues and work more effectively on the project.
If we can acquire a larger and more complete data set, we will be able to improve our system and potentially expand into other postural problems, including lordosis (curving inwards of the lower back), kyphosis (excessive curvature of the upper back), and scoliosis (sideways curving of the spine). Even beyond posture problems, our image analysis could be applied to other skin conditions and issues that can be diagnosed by a photograph.