add-biformer-and-add-fasternet-backbone-to-yolov5

cv-attention/readme.md

@@ -28,10 +28,11 @@
 | Polarized Self-Attention | True | https://arxiv.org/abs/2107.00782 |
 | Sequential Self-Attention | True | https://arxiv.org/abs/2107.00782 |
 | GAM | True | https://arxiv.org/pdf/2112.05561v1.pdf |
+| Biformer | True | https://arxiv.org/abs/2303.08810 |
 
 
 # Install
-部分注意力需要安装timm. 如运行提示缺少timm安装即可. 安装命令:pip install timm
+安装命令:pip install timm einops -i https://pypi.tuna.tsinghua.edu.cn/simple
 
 # Course
 1. [yolov5添加注意力哔哩哔哩视频教学链接](https://www.bilibili.com/video/BV1s84y1775U) [yolov5添加注意力-补充事项-哔哩哔哩视频教学链接](https://www.bilibili.com/video/BV1hG4y1M71X)
@@ -40,4 +41,5 @@
 
 # Reference
 https://github.com/xmu-xiaoma666/External-Attention-pytorch  
-https://github.com/rwightman/pytorch-image-models
+https://github.com/rwightman/pytorch-image-models
+https://github.com/rayleizhu/BiFormer

objectdetection-tricks/readme.md

@@ -14,4 +14,4 @@
     视频教学地址：[哔哩哔哩](https://www.bilibili.com/video/BV1LV4y1R7w6/).  
 - **tricks_4**
     用于yolov5和v7中的yolo格式转换coco格式的脚本.(如何在v5和v7中输出ap_small,ap_middle,ap_large coco指标)  
-    视频教学地址：[哔哩哔哩](https://www.bilibili.com/video/BV1LV4y1R7w6/).  
+    视频教学地址：[哔哩哔哩](https://www.bilibili.com/video/BV14T411s7Ts/).  

yolo-improve/readme.md

@@ -134,4 +134,8 @@
     添加Task-Specific Context Decoupling到yolov5中.  
     需要安装einops库. 命令: pip install -i https://pypi.tuna.tsinghua.edu.cn/simple einops  
     视频教学地址：[哔哩哔哩](https://www.bilibili.com/video/BV1mk4y1h7us/).   
-    paper: [yolov6链接](https://arxiv.org/pdf/2303.01047v1.pdf)  
+    paper: [yolov6链接](https://arxiv.org/pdf/2303.01047v1.pdf)  
+- **yolov5-backbone/fasternet**  
+    添加FasterNet主干到yolov5中.  
+    视频教学地址：[哔哩哔哩](https://www.bilibili.com/video/BV1Mx4y1A7jy/).   
+    reference: [链接](https://github.com/JierunChen/FasterNet)

yolo-improve/yolov5-backbone/fasternet/faster_cfg/fasternet_l.yaml

@@ -0,0 +1,13 @@
+mlp_ratio: 2
+embed_dim: 192
+depths: [3, 4, 18, 3]
+feature_dim: 1280
+patch_size: 4
+patch_stride: 4
+patch_size2: 2
+patch_stride2: 2
+layer_scale_init_value: 0 # no layer scale
+drop_path_rate: 0.3
+norm_layer:  BN
+act_layer: RELU
+n_div: 4

yolo-improve/yolov5-backbone/fasternet/faster_cfg/fasternet_m.yaml

@@ -0,0 +1,13 @@
+mlp_ratio: 2
+embed_dim: 144
+depths: [3, 4, 18, 3]
+feature_dim: 1280
+patch_size: 4
+patch_stride: 4
+patch_size2: 2
+patch_stride2: 2
+layer_scale_init_value: 0 # no layer scale
+drop_path_rate: 0.2
+norm_layer:  BN
+act_layer: RELU
+n_div: 4

yolo-improve/yolov5-backbone/fasternet/faster_cfg/fasternet_s.yaml

@@ -0,0 +1,13 @@
+mlp_ratio: 2
+embed_dim: 128
+depths: [1, 2, 13, 2]
+feature_dim: 1280
+patch_size: 4
+patch_stride: 4
+patch_size2: 2
+patch_stride2: 2
+layer_scale_init_value: 0 # no layer scale
+drop_path_rate: 0.1
+norm_layer:  BN
+act_layer: RELU
+n_div: 4

yolo-improve/yolov5-backbone/fasternet/faster_cfg/fasternet_t0.yaml

@@ -0,0 +1,13 @@
+mlp_ratio: 2
+embed_dim: 40
+depths: [1, 2, 8, 2]
+feature_dim: 1280
+patch_size: 4
+patch_stride: 4
+patch_size2: 2
+patch_stride2: 2
+layer_scale_init_value: 0 # no layer scale
+drop_path_rate: 0.
+norm_layer:  BN
+act_layer: GELU
+n_div: 4

yolo-improve/yolov5-backbone/fasternet/faster_cfg/fasternet_t1.yaml

@@ -0,0 +1,13 @@
+mlp_ratio: 2
+embed_dim: 64
+depths: [1, 2, 8, 2]
+feature_dim: 1280
+patch_size: 4
+patch_stride: 4
+patch_size2: 2
+patch_stride2: 2
+layer_scale_init_value: 0 # no layer scale
+drop_path_rate: 0.02
+norm_layer:  BN
+act_layer: GELU
+n_div: 4

yolo-improve/yolov5-backbone/fasternet/faster_cfg/fasternet_t2.yaml

@@ -0,0 +1,13 @@
+mlp_ratio: 2
+embed_dim: 96
+depths: [1, 2, 8, 2]
+feature_dim: 1280
+patch_size: 4
+patch_stride: 4
+patch_size2: 2
+patch_stride2: 2
+layer_scale_init_value: 0 # no layer scale
+drop_path_rate: 0.05
+norm_layer:  BN
+act_layer: RELU
+n_div: 4