Inspiration
We are often surrounded by disabled people. While they are granted access to many additional resources, it is often times impractical for someone to be by their side to translate what is going on in the sighted world. As children, we were always awestruck by the experiments done by CrazyRussianHacker, a youtuber who does interesting but simple science experiments. We hope that we may share that same joy with young visually impaired students by providing them with a convenient resource which helps them learn and keeps them safe.
What it does
The website lets you upload a photo of lab equipment and tells you the state and properties of the equipment.
How we built it
We built it through collecting over 6000 images from a kaggle dataset, processing them into tensors and feeding them into a resnet50 model with frozen layers and trained a dense layer at the top which provides the prediction output.
Challenges we ran into
Bias in the datasets and lack of sufficient data, lack of time to generate and train more, difficulty with compatibility in Sagemaker notebooks.
Accomplishments that we're proud of
Applying computer vision for a meaningful purpose and integrating our programming knowledge with AWS services for the first time.
What we learned
We learned how to use the vast amount of services AWS provides to help solve our ML/AI needs, as well as the importance of having good training data.
What's next for KDS
We aim to expand on the number of labels our model predicts, such that it can be used on a day-to-day basis!
Log in or sign up for Devpost to join the conversation.