Transformers AI Online Hackathon Summary

This project extracts text from images of product labels and uses GPT-3 to predict whether the substance is toxic to dogs. There are a number of potentially toxic substances to dogs, including food, rodenticides, plants, drugs (prescription and other), and mycotoxins (e.g., moldy grains). Since ingredients are important in determining the toxicity of a substance, the user can upload a photograph of the ingredient label or manually enter the ingredients. This can also be combined with external applications, e.g. for plant and mushroom identification. Disclaimer: this is a weekend hackathon project. If you think your pet may have been exposed to poison, please contact a professional and not GPT-3.

Poison

mEYE Buddy app is an application whose main role is to play part as your very own PERSONAL assistant who will make the world a little bit more accessible for anyone who is in any way visually impaired and requires assistance. The BEST SIDES OF THIS APP are that IT is highly affordable, with a premium version and is based on a loyalty program (business model). With mEYE buddy, not only would the life of blind people be much easier, but they could have the privilege of INDEPENDENCE - they would not need to rely on their caretaker or service dog to help them withq everyday chores.The state of the art tool for assistance to visually impaired persons! This state of the art AI technology will assist you in your everyday life. How app works: For first login you would have to register, and you'd do that using your fingerprint. Everytime the app is opened, the voice will tell you where everything is located on the screen. The design was made simple and the buttons were made big for easy access. The user can always press the icon in the middle of the screen for a reminder of the locations of the buttons. The voice assistant is the main function of the app. It will activate the AI, which will use a connected camera to describe the surroundings and warn the user of hazards. The key places tab will take the user to a tab where they can activate a guide to registered locations that are important to the user. (workplace, supermarket, hospital, home, coffee shop, etc.) They can also register new locations and remove old ones. The hazards that were noted by the ai could also be registered here. The devices tab is simple. It is used to connect the device to an external camera, preferably one on the glasses of the user or attached to their body. The user can also connect to their smart watch for easy access. The settings tab is pretty self-explanatory. However, besides the settings, it will also contain emergency information about the user, in case they need help from someone. The “mEYE Buddy” app will be connected to a camera on the glasses or on your body, and will describe everything important going in front of you - every peculiar movement that triggers its sensors or warn you against any potential hazards that may come in your way. The app can also be asked to describe something specific in more detail. Our Buddy also stores any hazard or newly recognized item in its already vast and enriched database. The app can also be told where certain points of interest are located (workplace, supermarket, hospital, home, coffee shop, etc.) for easier access to it later. The app comes with a simple UI with big buttons for input, and can be instructed through the AI voice as well. Also comes with a 3x3 keyboard for simpler accessibility.

Aurora

We have developed a Brain tumor classification app that helps doctors with the diagnosis. With this tool insurance and public health systems can launch massive brain tumor detection campaigns and save much time with the diagnosis, which will help in the early detection of this deadly disease.

LMF AI

Our Service is an aided diagnosis tool, that aids the workflow of doctors and also allows people to self-diagnose, both by means of AI

LastMinute

Imagine you're in some foreign city or country, this web app would help you navigate, socialize with locals, suggests you places to visit and make your stay way better.

Angelic

RecluAI is an api create with Django Rest Framework that uses the Co:Here api to generate Summaries and extract Information from interviews to help recruiter

Reclu-AI

Nowadays there are so many students and mothers who are willing to improve, but with no money (obviously with no work), are being unable to do so. Our team’s solution is this: Use the vacancy finder website while sorting it into the "knowledge" and directing it (via summarization through GPT-3) to the resources to learn from.

Artificial isnt artificial

The goal is to provide an interactive tool for visualizing research papers in a 2D map where the papers are placed on the map based on some vector embedding of the summary. The papers that are similar to each other will cluster together. This will allow researchers to identify papers that have similar concepts and even allow users to look at concepts from different fields that are applicable in their own field.

hummingbirds

The proposed solution alongside my teammates is to build a Fake News Detection Application meant to limit the propaganda effect. With the use of Node.js and GPT-3 trained model Davinci, we gave birth to our CrewBot .

TheCrew

This application is a chatbot for students, to reply for common asked questions in a certain material, we use cohere's model to answer students questions in a specific topic, the data used is WIKIFAQ and ideally emails from teachers of frequent questions and answers, the added value is that it's course specific and not like asking a broad question to google. I haven't advanced much in the implementation of the web interface and the model, but I wanted to submit the idea at least.

Student answer

Hello! We created this bot drawing inspiration from the COVID-19 pandemic in Singapore. Back in 2021, Singapore faced a surge of COVID-19 cases that overwhelmed the Ministry of Health in Singapore. Most people called the MOH hotline when they were unsure what to do with regard to COVID-19, but this flooded their hotlines and they were not able to answer most of their concerns. This was because they had no way to filter the urgency level of incoming messages and therefore had to answer to them one by one manually. We created this telegram bot to overcome this challenge. We made use of the text classification ability of GPT-3 to classify incoming messages to the bot based on their urgency levels, and the right level of assistance would then be provided.

Highway Volume

ChatBot

moose analytica