shape from x
play

Shape from X Haoqiang Fan fhq@megvii.com Some figures adapted from - PowerPoint PPT Presentation

Shape from X Haoqiang Fan fhq@megvii.com Some figures adapted from http://cvg.ethz.ch/teaching/2012spring/3dphoto/Slides/3dphoto12_shapeFromX.pdf Perception / Measurement of 3D 3D is vital for survival How to reconstruct / perceive 3D By


  1. Shape from X Haoqiang Fan fhq@megvii.com Some figures adapted from http://cvg.ethz.ch/teaching/2012spring/3dphoto/Slides/3dphoto12_shapeFromX.pdf

  2. Perception / Measurement of 3D 3D is vital for survival

  3. How to reconstruct / perceive 3D By means of visual information -> optical, 2D array of input

  4. Structure from Motion The most easy-to-understand approach Triangulation https://cn.mathworks.com/help/vision/ug/structure-from-motion.html

  5. Triangulation The epipolar constraint Stereo and kinect fusion for continuous 3D reconstruction and visual odometry

  6. Stereo, rectification, disparity row-to-row correspondence https://www.slideshare.net/DngNguyn43/stereo-vision-42147593

  7. Disparity, depth d=y_right - y_left z=B*F/d OpenCV: Depth Map from Stereo Images Middlebury Stereo Evaluation

  8. 3D Point Cloud x=x_screen/F*z y=y_screen/F*z Bundler: Structure from Motion (SfM) for Unordered Image Collections

  9. Surface Reconstruction Integration of oriented point

  10. Laplacian and Normal Laplacian = Normal * Mean Curvature

  11. SfM Scanning SLAM based positioning

  12. Depth Sensing: Active Sensors Structured Light Time of Flight(ToF)

  13. Structured Light Static pattern & dynamic pattern

  14. Time of Flight (ToF) Pulsed modulation

  15. Short Baseline Stereo Phase Detection Autofocus

  16. Shape from X Structure from Motion: 3D geometry Are there other possibilities?

  17. Shape from Shading Shading as a cue of 3D shape

  18. The Lambertian Law

  19. Shape from Shading Solve for gradient Assuming constant albedo

  20. Is Shape Uniquely Determined? bas-relief ambiguity

  21. Shape from Shading Data term + Prior

  22. Shape from Shading Example

  23. Photometric Stereo

  24. Photometric Stereo Measure the normal direction: the chrome sphere

  25. Depth from Normals

  26. Example Good for near Lambertian material

  27. Shape from Texture Solving normal from texture

  28. Depth from Focus Focus sweep

  29. Depth from Defocus Measure blur, solve depth

  30. Shape from Shadows Shadow carving 3D Reconstruction by Shadow Carving: Theory and Practical Evaluation”

  31. Shape from Specularities Solve deformation of mirrors. Toward a Theory of Shape from Specular Flow

  32. Shape from ? Shape from Nothing? Object priors!

  33. 3D Reconstruction from Single Image infer a whole shape, from a single image

  34. 3D Reconstruction from Single Image

  35. The ShapeNet Dataset

  36. 3D Reconstruction from Single Image

  37. 3D Reconstruction from Single Image

  38. The issue of representation

  39. Depth map

  40. Depth map

  41. Second depth map

  42. Second depth map

  43. The problem of discontinuity

  44. Volumetric Occupancy

  45. Problem of viewpoint

  46. Canonical View

  47. Volumetric Occupancy

  48. XML file

  49. XML file

  50. XML file

  51. XML file

  52. Can we find a representation that is.. flexible structural natural

  53. Point-based representation flexible structural natural

  54. Implementation details

  55. Results

  56. Results

  57. Results

  58. Human Performance

  59. A Neural Method to Stereo Matching

  60. Flownet & Dispnet Using raw left and right images as input Output disparity map End-to-End training

  61. Using two stacked images as input FlownetSimple

  62. Adding Correlation Layer Using correlation layer to explicitly provide cross view communication ability FlownetCorr

  63. Stereo Matching Cost Convolutional Neural Network Using CNN to calculate stereo matching cost between patches from different view Following with several post-process: Cross-based cost aggregation Semiglobal matching Left-right consistency check Disparity <-> Depth

  64. MRF Stereo methods We estimate f by minimizing the following energy function based on pairwise MRF Data term Smoothness term

  65. Global Local Stereo Neural Network Feature visualization

  66. results

  67. results

  68. results

  69. Implementation details Entangle two view feature inside network.

  70. Large Receptive Field Neural Network SimpleConv Encoder-Decoder SimpleConv simple conv ResConv blindingly increasing the receptive field of feature networks may not Improve the performance

  71. PatchMatch Communication Layer Directly provide the ability of communicating across two views

  72. Multi-staged Cascade

  73. Thanks Q/A

  74. 单击 以 结 束放映

  75. SemiGlobal Matching we define an energy function E(D) that depends on the disparity map D NP-Hard !!! But we can solve it through each directions to get an approximate solution by using Dynamic Programming(DP)

  76. Slanted patch matching The disparity d_p of each pixel p is over-parameterized by a local disparity plane Each pixels in the same plane has the same parameter (a_p, b_p, c_p) The true disparity maps are approximately piecewise linear We can estimate (a_p, b_p, c_p) for each pixel p instead of directly estimate d_p

Recommend


More recommend