EvalAI is an open source web application that helps researchers, students and data-scientists to create, collaborate and participate in various AI challenges organized round the globe.
In recent years, it has become increasingly difficult to compare an algorithm solving a given task with other existing approaches. These comparisons suffer from minor differences in algorithm implementation, use of non-standard dataset splits and different evaluation metrics. By providing a central leaderboard and submission interface, we make it easier for researchers to reproduce the results mentioned in the paper and perform reliable & accurate quantitative analysis. By providing swift and robust backends based on map-reduce frameworks that speed up evaluation on the fly, EvalAI aims to make it easier for researchers to reproduce results from technical papers and perform reliable and accurate analyses.
A question we’re often asked is: Doesn’t Kaggle already do this? The central differences are:
Custom Evaluation Protocols and Phases: We have designed versatile backend framework that can support user-defined evaluation metrics, various evaluation phases, private and public leaderboard.
Faster Evaluation: The backend evaluation pipeline is engineered so that submissions can be evaluated parallelly using multiple cores on multiple machines via mapreduce frameworks offering a significant performance boost over similar web AI-challenge platforms.
Portability: Since the platform is open-source, users have the freedom to host challenges on their own private servers rather than having to explicitly depend on Cloud Services such as AWS, Azure, etc.
Easy Hosting: Hosting a challenge is streamlined. One can create the challenge on EvalAI using the intuitive UI (work-in-progress) or using zip configuration file.
Centralized Leaderboard: Challenge Organizers whether host their challenge on EvalAI or forked version of EvalAI, they can send the results to main EvalAI server. This helps to build a centralized platform to keep track of different challenges.
Our ultimate goal is to build a centralized platform to host, participate and collaborate in AI challenges organized around the globe and we hope to help in benchmarking progress in AI.
Some background: Last year, the Visual Question Answering Challenge (VQA) 2016 was hosted on some other platform, and on average evaluation would take ~10 minutes. EvalAI hosted this year's VQA Challenge 2017. This year, the dataset for the VQA Challenge 2017 is twice as large. Despite this, we’ve found that our parallelized backend only takes ~130 seconds to evaluate on the whole test set VQA 2.0 dataset.
Setting up EvalAI on your local machine is really easy. You can setup EvalAI using two methods:
You can also use Docker Compose to run all the components of EvalAI together. The steps are:
Get the source code on to your machine via git.
git clone https://github.com/Cloud-CV/EvalAI.git evalai && cd evalai
Use your postgres username and password for fields
Build and run the Docker containers. This might take a while. You should be able to access EvalAI at
docker-compose up --build
Using Virtual Environment
Install python 2.7.10 or above, git, postgresql version >= 10.1, RabbitMQ and virtualenv, in your computer, if you don't have it already. If you are having trouble with postgresql on Windows check this link postgresqlhelp.
Get the source code on your machine via git.
git clone https://github.com/Cloud-CV/EvalAI.git evalai
Create a python virtual environment and install python dependencies.
cd evalai virtualenv venv source venv/bin/activate # run this command everytime before working on project pip install -r requirements/dev.txt
Create an empty postgres database.
sudo -i -u (username) createdb evalai
Change Postgresql credentials in
settings/dev.pyand run migrations
Use your postgres username and password for fields
dev.pyfile. After changing credentials, run migrations using the following command:
python manage.py migrate --settings=settings.dev
Seed the database with some fake data to work with.
python manage.py seed --settings=settings.dev
This command also creates a
host userand a
participant userwith following credentials.
HOST USER- username:
PARTICIPANT USER- username:
That's it. Now you can run development server at http://127.0.0.1:8000 (for serving backend)
python manage.py runserver --settings=settings.dev
Please make sure that node(
>=5.x.x) and bower(
>=1.8.x) are installed globally on your machine.
Install npm and bower dependencies by running
npm install bower install
If you running npm install behind a proxy server, use
npm config set proxy http://proxy:port
Now to connect to dev server at http://127.0.0.1:8888 (for serving frontend)
That's it, Open web browser and hit the url http://127.0.0.1:8888.
(Optional) If you want to see the whole game into play, then start the RabbitMQ worker in a new terminal window using the following command that consumes the submissions done for every challenge:
EvalAI is currently maintained by Deshraj Yadav, Akash Jain, Taranjeet Singh, Shiv Baran Singh and Rishabh Jain. A non-exhaustive list of other major contributors includes: Harsh Agarwal, Prithvijit Chattopadhyay, Devi Parikh and Dhruv Batra.
If you are interested in contributing to EvalAI, follow our contribution guidelines.