Diseases like ADHD, Autism, Cerebral Palsy, etc. have deep impacts on patients, irrespective of their age. Many movement disorders restrict a human’s movement, sense of balance, and posture, and these disorders have been very common ranging among children to adults. The most common interpretation of these studies says that “There is a high prevalence of substantial under-recognition and under-treatment of movement disorders in the general community”, as the current practices have been expensive and time-consuming relying on physical check-ups and the resultant conventional diagnosis. It is very important to diagnose Autism disorder at an early stage, as without a diagnosis this can make so many areas of life difficult, distressing, and bewildering for the undiagnosed person. Hence, we introduce Rollers an Autism detector software. which will alert parents as well as give brief information to doctors about abnormal activities.

What it does

It is a Cutting-Edge AI algorithm that will take a video of a child's movements as input and classifies what type of autism he/she is suffering from. At the same time, our Algorithm measures the rough distance of the child from the camera without sensors as well as it also measures by how much speed the movements are and in which direction it is moving using a state-of-the-art AI algorithm.

How we built it

We built it using Python, OpenCV, gluonCV, and MXNet.

Challenges we ran into

We ran through a lot of errors while coding the algorithm. At one moment it became totally difficult to train the algorithm, because it took 8 hours to train we need to speed it up to complete the project.

Accomplishments that we're proud of

Our algorithm is working!! given an input video of child it predicts what type of muscular disorder he is suffering from due to Autism.

What we learned

We learned tools like MXNet, OpenCV, and gluonCV.

What's next for Rollers

We would like to extend it for other diseases as well, like for ADHD, Cerebral palsy, etc.. and make our algorithm more robust.

Built With

Share this project: