v3.1.0 - YOLOv5 in PyTorch

The main body of this project comes from v7.0 - YOLOv5 SOTA Realtime Instance Segmentation of ultralytics/yolov5
This version uses GhostNet as the backbone network, and adds CBAM(Coordinate attention) attention mechanism in v3.0.0.

Important Updates

• Replacing the backbone network: CSPDarknet53 -> GhostNet, to make it lightweight🎉NEW
• Add CBAM attention mechanisms🎉NEW
  • After three Concats near the Head, 17/21/25Layer.
• Using VOC2007 dataset for training.

Specific Changes

• Add "models/parts_yolov5s-ghostnet_cbam.yaml"🎉NEW
  • After three Concats near the Head, 17/21/25Layer in "models/parts_yolov5s-ghostnet_cbam.yaml".
• Modify "train.py"
  • line 441: initial weights path -> weights/yolov5s.pt
  • line 442: model.yaml path -> models/parts_yolov5s-ghostnet_cbam🎉NEW
  • line 443: dataset.yaml path -> data/parts_voc2007.yaml

Notes

The modification of ECA and CA attention mechanisms is the same as CBAM.

v3.0.0 - YOLOv5 in PyTorch

The main body of this project comes from v7.0 - YOLOv5 SOTA Realtime Instance Segmentation of ultralytics/yolov5
This version uses GhostNet as the backbone network, and adds CA(Coordinate attention) attention mechanism in v2.0.0, placing it in 4 different positions.

Important Updates

• Replacing the backbone network: CSPDarknet53 -> GhostNet, to make it lightweight🎉NEW
• Add CA attention mechanisms and place them in 4 different positions🎉NEW
  • After three Concats near the Head, 17/21/25Layer.
  • After the first Concat, 13Layer.
  • Before SPPF, 9Layer.
  • Head last, 24Layer.
• Using VOC2007 dataset for training.

Specific Changes

• Modify "models/parts_yolov5s-ghostnet_ca(or ca2, ca3, ca4).yaml"🎉NEW
  • After three Concats near the Head, 17/21/25Layer in "models/parts_yolov5s-ghostnet_ca.yaml".
  • After the first Concat, 13Layer in "models/parts_yolov5s-ghostnet_ca2.yaml".
  • Before SPPF, 9Layer in "models/parts_yolov5s-ghostnet_ca3.yaml".
  • Head last, 24Layer in "models/parts_yolov5s-ghostnet_ca4.yaml".
• Modify "train.py"
  • line 441: initial weights path -> weights/yolov5s.pt
  • line 442: model.yaml path -> models/parts_yolov5s-ghostnet_ca(or ca2, ca3, ca4)🎉NEW
  • line 443: dataset.yaml path -> data/parts_voc2007.yaml

v2.3.0 - YOLOv5 in PyTorch

The main body of this project comes from v7.0 - YOLOv5 SOTA Realtime Instance Segmentation of ultralytics/yolov5.
This version adds SE attention mechanism in v2.0.0.

Important Updates

• Add SE attention mechanism to this network.🎉NEW
• Using VOC2007 dataset for training.

Specific Changes

• Modify "models/common_se.py"🎉NEW
  • Add Class: SEAttention.
• Modify "models/yolo.py"🎉NEW
  • Find parse_model function(line 303), adding SEAttention(line 322).
• Modify "train.py"
  • line 441: initial weights path -> weights/yolov5s.pt
  • line 442: model.yaml path -> models/parts_yolov5s_se.yaml🎉NEW
  • line 443: dataset.yaml path -> data/parts_voc2007.yaml

v2.2.0 - YOLOv5 in PyTorch

The main body of this project comes from v7.0 - YOLOv5 SOTA Realtime Instance Segmentation of ultralytics/yolov5.
This version adds ECA attention mechanism in v2.0.0.

Important Updates

• Add ECA attention mechanism to this network.🎉NEW
• Using VOC2007 dataset for training.

Specific Changes

• Modify "models/common_eca.py"🎉NEW
  • Add Class: ECAAttention.
• Modify "models/yolo.py"🎉NEW
  • Find parse_model function(line 303), adding ECAAttention(line 322).
• Modify "train.py"
  • line 441: initial weights path -> weights/yolov5s.pt
  • line 442: model.yaml path -> models/parts_yolov5s_eca.yaml🎉NEW
  • line 443: dataset.yaml path -> data/parts_voc2007.yaml

Notes

There are errors in "models/parts_yolov5s_eca.yaml", which will be modified in subsequent versions.⚠️ERROR

• layer 17: CoordAttention -> ECAAttention
• layer 21: CoordAttention -> ECAAttention
• layer 25: CoordAttention -> ECAAttention

v2.1.0 - YOLOv5 in PyTorch

The main body of this project comes from v7.0 - YOLOv5 SOTA Realtime Instance Segmentation of ultralytics/yolov5.
This version adds CBAM attention mechanism in v2.0.0.

Important Updates

• Add CBAM attention mechanism to this network.🎉NEW
• Using VOC2007 dataset for training.

Specific Changes

• Modify "models/common_cbam.py"🎉NEW
  • Add Class: ChannelAttention, SpatialAttention, CBAMBlock.
• Modify "models/yolo.py"🎉NEW
  • Find parse_model function(line 303), adding CBAMBlock(line 322).
• Modify "train.py"
  • line 441: initial weights path -> weights/yolov5s.pt
  • line 442: model.yaml path -> models/parts_yolov5s_cbam.yaml🎉NEW
  • line 443: dataset.yaml path -> data/parts_voc2007.yaml

v2.0.0 - YOLOv5 in PyTorch

The main body of this project comes from v7.0 - YOLOv5 SOTA Realtime Instance Segmentation of ultralytics/yolov5
This version adds CA(Coordinate attention) attention mechanism in v1.0.0, placing it in 4 different positions.

Important Updates

• Add CA attention mechanisms and place them in 4 different positions🎉NEW
  • After three Concats near the Head, 17/21/25Layer.
  • After the first Concat, 13Layer.
  • Before SPPF, 9Layer.
  • Head last, 24Layer.
• Using VOC2007 dataset for training.

Specific Changes

• Modify "models/parts_yolov5s_ca(or ca2, ca3, ca4).yaml"🎉NEW
  • After three Concats near the Head, 17/21/25Layer in "models/parts_yolov5s_ca.yaml".
  • After the first Concat, 13Layer in "models/parts_yolov5s_ca2.yaml".
  • Before SPPF, 9Layer in "models/parts_yolov5s_ca3.yaml".
  • Head last, 24Layer in "models/parts_yolov5s_ca4.yaml".
• Modify "train.py"
  • line 441: initial weights path -> weights/yolov5s.pt
  • line 442: model.yaml path -> models/parts_yolov5s_ca(or ca2, ca3, ca4)🎉NEW
  • line 443: dataset.yaml path -> data/parts_voc2007.yaml

v1.3.0 - YOLOv5 in PyTorch

The main body of this project comes from v7.0 - YOLOv5 SOTA Realtime Instance Segmentation of ultralytics/yolov5
This version adds the GhostNet on top of v1.0.0.
For GhostNet network structure, please refer to GhostNet: More Features from Cheap Operations.

Important Updates

• Replacing the backbone network: CSPDarknet53 -> GhostNet, to make it lightweight.🎉NEW
• Using VOC2007 dataset for training.

Specific Changes

• Due to the inclusion of GhostConv and C3Ghost modules required for GhostNet in YOLOv5/v7.0, there is no need to modify "models/common.py".🎉NEW
• Modify "train.py"
  • line 441: initial weights path -> weights/yolov5s.pt
  • line 442: model.yaml path -> models/parts_yolov5s-ghost.yaml🎉NEW
  • line 443: dataset.yaml path -> data/parts_voc2007.yaml

v1.2.0- YOLOv5 in PyTorch

The main body of this project comes from v7.0 - YOLOv5 SOTA Realtime Instance Segmentation of ultralytics/yolov5
This version adds the ShuffleNetV2 on top of v1.0.0.

Important Updates

• Replacing the backbone network: CSPDarknet53 -> ShuffleNetV2, to make it lightweight.🎉NEW
• Using VOC2007 dataset for training.

Specific Changes

• Modify "models/common.py"🎉NEW
  • Add modules required for ShuffleNetV2.
  • Add Class: ConvBNReLuMaxpool, ShuffleBlock.
• Modify "models/yolo.py"🎉NEW
  • Find parse_model function(line 303), adding ConvBNReLuMaxpool, ShuffleBlock(line 322).
• Modify "train.py"
  • line 441: initial weights path -> weights/yolov5s.pt
  • line 442: model.yaml path -> models/parts_yolov5s_shufflenetv2.yaml🎉NEW
  • line 443: dataset.yaml path -> data/parts_voc2007.yaml

v1.1.2 - YOLOv5 in PyTorch

The main body of this project comes from v7.0 - YOLOv5 SOTA Realtime Instance Segmentation of ultralytics/yolov5
This version adds the MobileNetV3 on top of v1.1.1.

Important Updates

• Replacing the backbone network: CSPDarknet53 -> MobileNetV3-Large, to make it lightweight.🎉NEW
• Using VOC2007 dataset for training.

Specific Changes

• Modify "models/common.py"
  • Add modules required for MobileNetV3-Large.
  • Add Class: h_sigmoid, h_swish, SELayer, conv_bn_hswish, InvertedResidual3(Different from InvertedResidual in MobileNetV2, so renamed InvertedResidual3).
• Modify "models/yolo.py"
  • Find parse_model function(line 303), adding h_sigmoid, h_swish, SELayer, conv_bn_hswish, InvertedResidual3(line 322).
• Modify "train.py"
  • line 441: initial weights path -> weights/yolov5s.pt
  • line 442: model.yaml path -> models/parts_yolov5s_mobilenetv3l.yaml🎉NEW
  • line 443: dataset.yaml path -> data/parts_voc2007.yaml

v1.1.1- YOLOv5 in PyTorch

The main body of this project comes from v7.0 - YOLOv5 SOTA Realtime Instance Segmentation of ultralytics/yolov5
This version adds the MobileNetV3 on top of v1.0.0.

Important Updates

• Replacing the backbone network: CSPDarknet53 -> MobileNetV3-Small, to make it lightweight.🎉NEW
• Using VOC2007 dataset for training.

Specific Changes

• Modify "models/common.py"🎉NEW
  • Add modules required for MobileNetV3-Small.
  • Add Class: h_sigmoid, h_swish, SELayer, conv_bn_hswish, InvertedResidual3(Different from InvertedResidual in MobileNetV2, so renamed InvertedResidual3).
• Modify "models/yolo.py"🎉NEW
  • Find parse_model function(line 303), adding h_sigmoid, h_swish, SELayer, conv_bn_hswish, InvertedResidual3(line 322).
• Modify "train.py"
  • line 441: initial weights path -> weights/yolov5s.pt
  • line 442: model.yaml path -> models/parts_yolov5s_mobilenetv3s.yaml🎉NEW
  • line 443: dataset.yaml path -> data/parts_voc2007.yaml