--- tags: 2021-UnityRL --- # Experimental Results ## 2D motion environment ![](https://i.imgur.com/Ka3U8kY.png =60%x) ### Baselines * Optical Flow * Optical Flow w. Instance * Semantic Segmentation (binary mask, motion / no-motion) * Semantic Segmentation (multi-class mask, person / car / road ...) * w/ and w/o depth ## 3D motion environment ![](https://i.imgur.com/5trqWJ1.png =60%x) * 2D Optical Flow * 2D Optical Flow w. Instance * 3D Optical Flow (Scene Flow) * 3D Optical Flow (Scene Flow) w. Instance (Better Compare with Rigid Flow / Non-Rigid Flow?) * Semantic Segmentation (binary mask, motion / no-motion) * Semantic Segmentation (multi-class mask, person / car / road ...)