Home Explore Help
Sign In
SRI-DYZBC2
/
Vehicle-cpp
2
0
Fork 0
Files Issues 0 Pull Requests 0 Wiki
Branch: master
Branches Tags
Vehicle-cpp
/
hkpc
/
vision
/
docs
/
source
/
ops.rst

ops.rst 1.8 KB
Permalink History Raw

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103 
  1. .. _ops:
    2. 

  2. 

  3. Operators
    4. 

  4. =========
    5. 

  5. 

  6. .. currentmodule:: torchvision.ops
    7. 

  7. 

  8. :mod:`torchvision.ops` implements operators, losses and layers that are specific for Computer Vision.
    9. 

  9. 

  10. .. note::
    11. 

  11.   All operators have native support for TorchScript.
    12. 

  12. 

  13. 

  14. Detection and Segmentation Operators
    15. 

  15. ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
    16. 

  16. 

  17. The below operators perform pre-processing as well as post-processing required in object detection and segmentation models.
    18. 

  18. 

  19. .. autosummary::
    20. 

  20.     :toctree: generated/
    21. 

  21.     :template: function.rst
    22. 

  22. 

  23.     batched_nms
    24. 

  24.     masks_to_boxes
    25. 

  25.     nms
    26. 

  26.     roi_align
    27. 

  27.     roi_pool
    28. 

  28.     ps_roi_align
    29. 

  29.     ps_roi_pool
    30. 

  30. 

  31. .. autosummary::
    32. 

  32.     :toctree: generated/
    33. 

  33.     :template: class.rst
    34. 

  34. 

  35.     FeaturePyramidNetwork
    36. 

  36.     MultiScaleRoIAlign
    37. 

  37.     RoIAlign
    38. 

  38.     RoIPool
    39. 

  39.     PSRoIAlign
    40. 

  40.     PSRoIPool
    41. 

  41. 

  42. 

  43. Box Operators
    44. 

  44. ~~~~~~~~~~~~~
    45. 

  45. 

  46. These utility functions perform various operations on bounding boxes.
    47. 

  47. 

  48. .. autosummary::
    49. 

  49.     :toctree: generated/
    50. 

  50.     :template: function.rst
    51. 

  51. 

  52.     box_area
    53. 

  53.     box_convert
    54. 

  54.     box_iou
    55. 

  55.     clip_boxes_to_image
    56. 

  56.     complete_box_iou
    57. 

  57.     distance_box_iou
    58. 

  58.     generalized_box_iou
    59. 

  59.     remove_small_boxes
    60. 

  60. 

  61. Losses
    62. 

  62. ~~~~~~
    63. 

  63. 

  64. The following vision-specific loss functions are implemented:
    65. 

  65. 

  66. .. autosummary::
    67. 

  67.     :toctree: generated/
    68. 

  68.     :template: function.rst
    69. 

  69. 

  70.     complete_box_iou_loss
    71. 

  71.     distance_box_iou_loss
    72. 

  72.     generalized_box_iou_loss
    73. 

  73.     sigmoid_focal_loss
    74. 

  74. 

  75. 

  76. Layers
    77. 

  77. ~~~~~~
    78. 

  78. 

  79. TorchVision provides commonly used building blocks as layers:
    80. 

  80. 

  81. .. autosummary::
    82. 

  82.     :toctree: generated/
    83. 

  83.     :template: class.rst
    84. 

  84. 

  85.     Conv2dNormActivation
    86. 

  86.     Conv3dNormActivation
    87. 

  87.     DeformConv2d
    88. 

  88.     DropBlock2d
    89. 

  89.     DropBlock3d
    90. 

  90.     FrozenBatchNorm2d
    91. 

  91.     MLP
    92. 

  92.     Permute
    93. 

  93.     SqueezeExcitation
    94. 

  94.     StochasticDepth
    95. 

  95. 

  96. .. autosummary::
    97. 

  97.     :toctree: generated/
    98. 

  98.     :template: function.rst
    99. 

  99. 

  100.     deform_conv2d
    101. 

  101.     drop_block2d
    102. 

  102.     drop_block3d
    103. 

  103.     stochastic_depth
    104.