Aspergers

• 

  Ad Model as an offset of Intelligent Models in Autonomously (perhaps Intelligent) Driven Cars

Inspiration

Building an ad service model as a subset of self-driving car model lidar and optical camera inferencing.

What it does

We know that self-driven cars have LIDAR and optical camera for perception and navigation tracking purposes. Idea is to use those models to build an ad business model in which when we encounter shops/ retails on the way which pays us (subscribed to us), we can induce the driver (user) to shop in those shops by giving details on current discounts and features available in those shops.

How we built it

For prototype, we trained a PyTorch model with Amazon E-Commerce catalog We experimented Inception-V3 CNN models as out backbone and built a model to identify products based on Images

Challenges we ran into

Data collection was the primary challenge we ran into as we had to Engineer our data making multiple joints from different sources in-order to create a dataset viable to be trained. We also ran into several CUDA related issues since we were forced into using GPU as we trained our model in a pretty big set (about 10^5 images) and it would not have been feasible to train it in a CPU based system.

Accomplishments that we're proud of

We were able to build a working model (although pretty simple) within stipulated time trained on Amazon E-Commerce Dataset. We followed industry standards when building our Dataset and DataLoader modules.

What we learned

Few technical intricacies related to data loading aspects in PyTorch-CUDA configs.

What's next for Aspergers

We will try to make this into a viable product in market. Our immediate next goal will be to build a voice command NLP system to notify the users of possible shopping destinations during their travel.

Built With

• pandas
• pytorch