A good vision is one of the key ingredients for building an AI. Although we see the power of deep learning and convolutional neural networks everyday in various platforms, we thought it would be nice to see how an embodied AI does when given tasks that require a strong object detection performance.
What it does
When faced with various objects, the GoPiGo robot listens to a user's object of choice, detects it, and moves toward that object.
How we built it
We used the rigorous pre-trained CNN model designed at Google called Inception to detect images and their locations.
Challenges we ran into
Setting up GoPiGo and fine tuning the model with custom hyper parameters as we saw fit.
What's next for Envision
Publishing the project as open source.