On March 16–18, Dbrain together with Open Data Science and Mail.Ru Group hosted Whatever Hack — a machine learning hackaton in Moscow with a prize fund of 1.5 million rubles.
180 participants were creating solutions in three nominations.
It’s not funny
The first one called was to identify if a description matches a picture — in other words, to evaluate humor quality in memes. The winning team got a gradient boosting over TF-IDF on captions and tags.
Two chairs
The next was to train a furniture detector. The team of Connecthome.ai and Constanta employees came first. Cracking spaghetti-code ‘keras’ made by Mask-RCNN was enough to overcome Faster RCNN’s one with their mxnet-based answer.
Whatever
A winning team in this nomination created an important social service. The participants took a poorly designed vacancy, which caused a lot of negative feedback from ODS users, analyzed key requirements and generated fake CVs based on the verified ODS users’ profiles. After reaching the critical level of complaints, all the CVs were sent to the dishonest employer.
Additionally, there was a team with a great idea to create meme recommendations based on the users’ emotions recognition via a frontal smartphone camera.
It was only a start, stay tuned for more events by Dbrain.
Join the Telegram chat, follow us on Twitter, like us on Facebook
