Simulated Visual Navigation Dataset for "Two-Stage Visual Navigation by Deep Neural Networks and Multi-Goal Reinforcement Learning"

Condensed data set (train, validation, test) of two simulated environments in the Gazebo simulator. One small kitchen of approximately 45 square meters and a big room of roughly 140 square meters. The data can be loaded using numpy. The images are size 84x84x3. The labels for each image are (X position, Y position, sine of theta, and cosine of theta).