The paper "DeepDriving: Learning Affordance for Direct Perception in Autonomous Driving" by Chenyi Chen, Ari Seff, Alain Kornhauser, and Jianxiong Xiao proposes a novel paradigm for autonomous driving systems that lies between mediated perception and behavior reflex approaches. The authors introduce a direct perception approach to estimate the affordance for driving, mapping an input image to a small set of key perception indicators related to road/traffic states. This representation provides a compact yet complete description of the scene, enabling a simple controller to make autonomous driving decisions. The model is trained using 12 hours of human driving data from the video game TORCS and demonstrates good performance in diverse virtual environments. Additionally, the authors train a model for car distance estimation on the KITTI dataset, showing that their direct perception approach generalizes well to real driving images. The paper includes a detailed description of the system architecture, training process, and evaluation results, highlighting the effectiveness of the proposed approach in both virtual and real-world scenarios.The paper "DeepDriving: Learning Affordance for Direct Perception in Autonomous Driving" by Chenyi Chen, Ari Seff, Alain Kornhauser, and Jianxiong Xiao proposes a novel paradigm for autonomous driving systems that lies between mediated perception and behavior reflex approaches. The authors introduce a direct perception approach to estimate the affordance for driving, mapping an input image to a small set of key perception indicators related to road/traffic states. This representation provides a compact yet complete description of the scene, enabling a simple controller to make autonomous driving decisions. The model is trained using 12 hours of human driving data from the video game TORCS and demonstrates good performance in diverse virtual environments. Additionally, the authors train a model for car distance estimation on the KITTI dataset, showing that their direct perception approach generalizes well to real driving images. The paper includes a detailed description of the system architecture, training process, and evaluation results, highlighting the effectiveness of the proposed approach in both virtual and real-world scenarios.