Deep Vision was created to address the democratized use of AI. A lot of services are trying to address that but we consider that they are only scraping the surface of that idea. For AI to be democratized, everyone should be able to use it regardless of their background and income level. That means that they do not need to have the technical knowledge nor the money to hire expensive consulting companies.
Deep Vision is a suite of deep learning products that enable companies and individuals with limited machine learning expertise to train high-quality models specific to their business needs. We aim to bring AI to unattended sectors around the world.
The key factors that make this product unique are the following:
-
The Hub: A space where people can create, share and collaborate with-in projects around the world without a need of prior technical knowledge (Basic/Premium plan).
-
Performance adapted to needs: Auto-ML or the help of qualified professionals? What suits best your needs? Customers can choose between multiple options according to their needs (Basic/Premium plan).
-
Easy of use: This service aims to bridge the gap between State Of The Art models and the general public. For those reasons, one of our main imperatives is ease of use.
To that end, we have created a web and a mobile applications for the service.
| Web Application | Mobile Application |
|---|
|
After logging in and subscribing for a specific plan a user can start creating projects. A project is meant to be used for a specific task. For instance, classifying types of fruits.
We break the whole process into three main components.
-
Upload: Every Machine Learning algorithm is based on data. While we can bring the technical expertise we require the user to bring the labelled data. The more data the better the performance.
-
Train: When the project has enough data the user can decide to train the model. We will send an email when the training has finished with information about the current performance. Then, they can decide if they want to keep collecting data or they require the services of our professionals.
-
Evaluate: Once the model has been trained the user can now use it. They can do so through the mobile app or through the web app depending on their needs. If they need predictions in the field we recommend the mobile app while the web app is better if you want, for example, to classify data that is in your computer.
Once the project has been created the owner can add collaborators and they can start collecting data or using the model at the same time.
Let's go through the first bit of the user-flow. When a user logs-in, the first thing they will see is the list of projects that they have or they are collaborating in.
When a project is clicked they can, then, choose what they want to do: Upload it, train it or use it.
You may think that it can get complex for the user. Let's look at one of the functionalities to see if that might be the case. Let's choose to upload for example.
We believe that the project is still far from being complete. Some of the components we would like to integrate on the future are:
We also think that there are additional benefits to that project. If an individual or an organization collect a reach and variate data-set we could be an intermediary between the AI community and the individual to benefit both through commerce.
This project is still under development. If you are seriously interested in contributing to the idea please contact me. My name is Robert and you can write to me at risandor@usfca.edu.
This project has been develop through a team effort of:
Miguel Romero, Robert Sandor, Hai Vu Le, Liying Li, Meng-Ting Chang, Zhi Li and Wendy Xiao.