A Machine Learning Approach to Visual Perception of Forest Trails for Mobile Robots

This dataset is a part of the supplementary materials to the 2017 RAL article with the same title. A Machine Learning Approach to Visual Perception of Forest Trails for Mobile Robots IEEE Robotics and Automation Letters Alessandro Giusti, Jerome Guzzi, Dan Ciresan, Fang Lin He, Juan Pablo Rodriguez, Flavio Fontana, Matthias Faessler, Christian Forster, Jurgen Schmidhuber, Gianni A. Di Caro, Davide Scaramuzza, Luca Gambardella You can find more information on the project web page (alessandrog@idsia.ch). Dataset Folders 001..010 contain the dataset used to train the networks. Folder 000 contains preliminary test data. Folders 011..014 contain data for testing the system. 000 and 003 were shot with an handheld cellphone. 001 and 002 were shot with 3 GOPRO Hero 3 cameras, fixed on the head with straps. 004..014 were shot with 3 Bluefox cameras, fixed on a rigid helm (the same model and with the same lens as the camera mounted on the quadcopter)