Results of the second Daisi Hackathon
The second Daisi Hackathon just finished and we are pleased to annouce the winners!
Judge’s Choice prizes — Best Daisies
We have received over 70 high quality and very competitive submissions. Here below are the three winners.
- First prize — Automatic Cheque Clearing System (Chirag Hedge)
In spite of the overall rapid emergence of electronic payments, huge volumes of handwritten bank cheques are issued and processed manually every day.
Automatic Cheque Clearing System wraps a very elaborate and well documented network of Daisies, illustrating the mechanisms of Daisi chaining and how easy it is to build a reliable microservices backend with the Daisi platform.
The Automatic Cheque Clearing System leverages AI fuelled technology along with rich rule-based automations to optimize, organize and streamline cheque clearance.
- Second prize — Text Imager (Anshu Kumar)
Diffussion based models are the next trend in computer vision. This Daisi is the sophisticated deployment in a web app of many image or video generation models, taking full advantage of GPUs as a service on the Daisi platform.
In addition to generating images on demanded prompts with instant speed thanks to platform, It can generate similar images using Img2Img section with some additional features like masking adding emotions etc. It can also generate videos by combining multiple frames generated using Stable Diffusion. And it can generate captions from a given image using blip caption model and clip models like ViT-L/14, ViT-H-14 and others .
The app features also a concept library with over 500 image styles trained with Textual Inversion getting more updates.
- Third prize Doc2MD (Sadiki Abdeladim)
Digitalization of documents is a hard problem to solve, which requires to combine many different techniques. This Daisi wraps elegantly a series of models deployed as services on the Daisi platform and proposes an efficient pipeline with Daisi chaining to extract value from PDF documents by converting their content to a Markdown format.
Most called Daisi prizes
Daisies as microservices are to be called from third party systems. The “Most called Daisi” prize is awarded to the submitted Daisies which have received the most API calls during the hackathon period.
The wining submissions are the following
- First prize Whisper Model GPU (Kanav Anad )
Whisper is an automatic speech recognition model trained on 680,000 hours of multilingual data collected from the web. As per OpenAI, the model is robust to accents, background noise and technical language. In addition, it supports 99 different languages’ transcription and translation from those languages into English. This Daisi is the deployment of the small model and has been called more than 15.6M times during the hackathon period.
- Second prize - Iris Flower Prediction App (Nandhakumar Shanmugam)
The Iris flowers classification problem is a classic in machine learning, solved here with a Random Forest algorithm and deployed as a Daisi with a nice UI. It has been called over 5.3M times during the hackathon period.
- Third prize — Analysis of cryptocurrencies (Mohamed Farhun)
Cryptocurrencies play a growing role in modern economy. This Daisi provides useful metrics to analyse markets and has been called 1.5M times during the hackathon period.
Congratulations to all winners and to every participant. Given the overall excellent quality of the submissions, selecting the winning Daisies has been a difficult exercise.
The Daisi platform will be shut down on November 15 (see blog post here). We are immensely grateful to everyone, developers, users, visitors, who have made possible to bring it this far. 1