SynthMining

Inspiration

Currently, one of the largest issues in image-based data science - specifically image generation and machine learning - is the lack of large amounts of data necessary for data mining, and the time necessary to train them. To provide a solution, we created an online platform that allows collaborative outsourced data mining that connects contributors to those with immature algorithms.

What it does

Such contributors are paid to submit valid data samples to those who need data sets to train machine learning via a built-in Generating Adversarial Network (GAN). Those wising to train networks are able to search by tags in order to find the right data sets needed, while data providers and paid to submit valid data sets, e.g. images. The project consists of two main parts, the GAN itself, and the platform it is hosted on.

Challenges we ran into

Tackling machine learning and creating a project around it was definitely a struggle, as most of us had little prior experience with understanding the concepts and applying it to code in a 24-hour timeframe.

Accomplishments that we're proud of

Despite the complexity of the subject we chose, our ambition paid off and we managed to complete the

What we learned

We learned that the incredible versatility of code, and that coding is ultimately an application of the knowledge that others have put out, with a myriad of powerful results.

What's next for SynthMining

SynthMining still has several areas that could be refined, and polished, but the next long term plan is to increase its accessibility and appeal. The quality and benefits of using the platform would only increase as time continues, and in this lies the next step to breaking through one of the roadblocks facing data scientists today.