rtdetr-x.yaml 2.2 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657
  1. # Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license
  2. # Ultralytics RT-DETR-x hybrid object detection model with P3/8 - P5/32 outputs
  3. # Model docs: https://docs.ultralytics.com/models/rtdetr
  4. # Task docs: https://docs.ultralytics.com/tasks/detect
  5. # Parameters
  6. nc: 80 # number of classes
  7. scales: # model compound scaling constants, i.e. 'model=yolov8n-cls.yaml' will call yolov8-cls.yaml with scale 'n'
  8. # [depth, width, max_channels]
  9. x: [1.00, 1.00, 2048]
  10. backbone:
  11. # [from, repeats, module, args]
  12. - [-1, 1, HGStem, [32, 64]] # 0-P2/4
  13. - [-1, 6, HGBlock, [64, 128, 3]] # stage 1
  14. - [-1, 1, DWConv, [128, 3, 2, 1, False]] # 2-P3/8
  15. - [-1, 6, HGBlock, [128, 512, 3]]
  16. - [-1, 6, HGBlock, [128, 512, 3, False, True]] # 4-stage 2
  17. - [-1, 1, DWConv, [512, 3, 2, 1, False]] # 5-P3/16
  18. - [-1, 6, HGBlock, [256, 1024, 5, True, False]] # cm, c2, k, light, shortcut
  19. - [-1, 6, HGBlock, [256, 1024, 5, True, True]]
  20. - [-1, 6, HGBlock, [256, 1024, 5, True, True]]
  21. - [-1, 6, HGBlock, [256, 1024, 5, True, True]]
  22. - [-1, 6, HGBlock, [256, 1024, 5, True, True]] # 10-stage 3
  23. - [-1, 1, DWConv, [1024, 3, 2, 1, False]] # 11-P4/32
  24. - [-1, 6, HGBlock, [512, 2048, 5, True, False]]
  25. - [-1, 6, HGBlock, [512, 2048, 5, True, True]] # 13-stage 4
  26. head:
  27. - [-1, 1, Conv, [384, 1, 1, None, 1, 1, False]] # 14 input_proj.2
  28. - [-1, 1, AIFI, [2048, 8]]
  29. - [-1, 1, Conv, [384, 1, 1]] # 16, Y5, lateral_convs.0
  30. - [-1, 1, nn.Upsample, [None, 2, "nearest"]]
  31. - [10, 1, Conv, [384, 1, 1, None, 1, 1, False]] # 18 input_proj.1
  32. - [[-2, -1], 1, Concat, [1]]
  33. - [-1, 3, RepC3, [384]] # 20, fpn_blocks.0
  34. - [-1, 1, Conv, [384, 1, 1]] # 21, Y4, lateral_convs.1
  35. - [-1, 1, nn.Upsample, [None, 2, "nearest"]]
  36. - [4, 1, Conv, [384, 1, 1, None, 1, 1, False]] # 23 input_proj.0
  37. - [[-2, -1], 1, Concat, [1]] # cat backbone P4
  38. - [-1, 3, RepC3, [384]] # X3 (25), fpn_blocks.1
  39. - [-1, 1, Conv, [384, 3, 2]] # 26, downsample_convs.0
  40. - [[-1, 21], 1, Concat, [1]] # cat Y4
  41. - [-1, 3, RepC3, [384]] # F4 (28), pan_blocks.0
  42. - [-1, 1, Conv, [384, 3, 2]] # 29, downsample_convs.1
  43. - [[-1, 16], 1, Concat, [1]] # cat Y5
  44. - [-1, 3, RepC3, [384]] # F5 (31), pan_blocks.1
  45. - [[25, 28, 31], 1, RTDETRDecoder, [nc]] # Detect(P3, P4, P5)