An Unexpected Journey to Сomputer Vision and Object Detection, or There and Back Again

During the talk, we will delve into the world of image recognition on mobile platforms. The speaker and his team developed a real-time map displaying road events, using video data analysis from dashcams. One of the key challenges involved creating and implementing new, more efficient models for object detection in the video stream.

Anton will introduce the fundamentals of computer vision, explore available frameworks for creating and training object detection models, and highlight the advantages of the team's approach, specifically the usage of YOLO models and the TensorFlow library on Android. He will also share real-world challenges – ranging from difficulties to unexpected errors the team encountered along the way.