Kirill Romanov AutoCV2

Kirill Romanov tells about his participation in AutoCV2 challenge in English. His solution got third place. In this video you will find out: - Some tips and tricks for models’ speed training from scratch within limited time and space (hardware capacities) budget - How to adapt picture augmentations to a specific dataset on the fly - How to build a simple but efficient picture classification system that works surprisingly well across a different picture and video datasets - The limitations and areas of improvement for the solution Presentation - Find out about new competitions Find out about new machine learning training: VKontakte Facebook Telegram