You can not select more than 25 topics Topics must start with a chinese character,a letter or number, can include dashes ('-') and can be up to 35 characters long.

rpn_ops.h 2.1 kB

5 years ago
3 years ago
5 years ago
5 years ago
5 years ago
5 years ago
12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061
  1. /**
  2. * Copyright 2019 Huawei Technologies Co., Ltd
  3. *
  4. * Licensed under the Apache License, Version 2.0 (the "License");
  5. * you may not use this file except in compliance with the License.
  6. * You may obtain a copy of the License at
  7. *
  8. * http://www.apache.org/licenses/LICENSE-2.0
  9. *
  10. * Unless required by applicable law or agreed to in writing, software
  11. * distributed under the License is distributed on an "AS IS" BASIS,
  12. * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
  13. * See the License for the specific language governing permissions and
  14. * limitations under the License.
  15. */
  16. /*!
  17. * \file rpn_ops.h
  18. * \brief
  19. */
  20. #ifndef OPS_BUILT_IN_OP_PROTO_INC_RPN_OPS_H_
  21. #define OPS_BUILT_IN_OP_PROTO_INC_RPN_OPS_H_
  22. #include "graph/operator_reg.h"
  23. namespace ge {
  24. /**
  25. *@brief Iteratively removes lower scoring boxes which have an IoU greater than
  26. * iou_threshold with higher scoring box according to their
  27. * intersection-over-union (IoU) . \n
  28. *@par Input:
  29. * @li box_scores: 2-D tensor with shape of [N, 8], including proposal boxes and
  30. * corresponding confidence scores . \n
  31. * @par Attributes:
  32. * @li iou_threshold: An optional float. The threshold for deciding whether boxes
  33. * overlap too much with respect to IOU . \n
  34. * @par Outputs:
  35. * @li selected_boxes: 2-D tensor with shape of [N,5], representing filtered
  36. * boxes including proposal boxes and corresponding confidence scores.
  37. * @li selected_idx: 1-D tensor with shape of [N], representing the index of
  38. * input proposal boxes.
  39. * @li selected_mask: 1-D tensor with shape of [N], the symbol judging whether
  40. * the output proposal boxes is valid . \n
  41. * @attention Constraints:
  42. * The 2nd-dim of input box_scores must be equal to 8.\n
  43. * Only supports 2864 input boxes at one time.\n
  44. */
  45. REG_OP(NMSWithMask)
  46. .INPUT(box_scores, TensorType({DT_FLOAT, DT_FLOAT16}))
  47. .OUTPUT(selected_boxes, TensorType({DT_FLOAT, DT_FLOAT16}))
  48. .OUTPUT(selected_idx, TensorType({DT_INT32}))
  49. .OUTPUT(selected_mask, TensorType({DT_UINT8}))
  50. .ATTR(iou_threshold, Float, 0.5)
  51. .OP_END_FACTORY_REG(NMSWithMask)
  52. } // namespace ge
  53. #endif // OPS_BUILT_IN_OP_PROTO_INC_RPN_OPS_H_

图引擎模块(GE)是MindSpore的一个子模块,其代码由C++实现,位于前端模块ME和底层硬件之间,起到承接作用。图引擎模块以ME下发的图作为输入,然后进行一系列的深度图优化操作,最后输出一张可以在底层硬件上高效运行的图。GE针对昇腾AI处理器的硬件结构特点,做了特定的优化工作,以此来充分发挥出昇腾AI处理器的强大算力。在进行模型训练/推理时,GE会被自动调用而用户并不感知。GE主要由GE API和GE Core两部分组成,详细的架构图如下所示