lyuwenyu / Geometry

Depth Prediction, Optical Flow, Pose Estimation. etc.

Depth prediction

Depth prediction

---

self-supervised monocular

• Assumptions

moving camera, static scene, rigid, appearance, Lambertian

• Issues

out-of-view, occlusion, moving objects, stationary, low-texture

---

0. Moving Camera, Moving People: A Deep Learning Approach to Depth Prediction
1. Digging into Self-Supervised Monocular Depth Prediction-monodepth2
2. Unsupervised Learning of Depth and Ego-Motion from video
3. Unsupervised Monocular Depth Estimation with Left-Right Consistency-monodepth
4. Depth prediction without the sensors: Leveraging structure for unsupervised learning-struct2depth
5. GeoNet: Unsupervised Learning of Dense Depth, Optical Flow and Camera Pose

---

Optical Flow

Optical Flow

The 2D optical flow between each input frame and another frame in the video, which represents the pixel displacement between the two frames. This flow field depends on both the scene’s depth and the relative position of the camera.

1. Self-Supervised Learning via Conditional Motion Propagation