search

Gumpest / YOLOv5-Multibackbone-Compression

YOLOv5 Series Multi-backbone(TPH-YOLOv5, Ghostnet, ShuffleNetv2, Mobilenetv3Small, EfficientNetLite, PP-LCNet, SwinTransformer YOLO), Module(CBAM, DCN), Pruning (EagleEye, Network Slimming), Quantization (MQBench) and Deployment (TensorRT, ncnn) Compression Tool Box.
994 stars 200 forks source link

SwinTrans-YOLOv5它的预权重可以是yolov5s或者yolov5m吗 #62

Open bjw1107 opened 2 years ago

zaipinai commented 2 years ago

depth_multiple: 1.33 # model depth multiple width_multiple: 1.25 # layer channel multiple 改下模型文件这俩行就能用了

bjw1107 commented 2 years ago

谢谢帮助🙏

发自我的iPhone

------------------ 原始邮件 ------------------ 发件人: zaipinai @.> 发送时间: 2022年3月13日 20:08 收件人: Gumpest/YOLOv5-Multibackbone-Compression @.> 抄送: bjw1107 @.>, Author @.> 主题: Re: [Gumpest/YOLOv5-Multibackbone-Compression] SwinTrans-YOLOv5它的预权重可以是yolov5s或者yolov5m吗 (Issue #62)

depth_multiple: 1.33 # model depth multiple width_multiple: 1.25 # layer channel multiple 改下模型文件这俩行就能用了

— Reply to this email directly, view it on GitHub, or unsubscribe. Triage notifications on the go with GitHub Mobile for iOS or Android. You are receiving this because you authored the thread.Message ID: @.***>

PeterSmith1 commented 2 years ago

depth_multiple: 1.33 # model depth multiple width_multiple: 1.25 # layer channel multiple 改下模型文件这俩行就能用了

请问一下，你使用这个yaml训练结果是多少阿

bjw1107 commented 2 years ago

用的自己做的数据集，跑出来效果一般般

发自我的iPhone

------------------ 原始邮件 ------------------ 发件人: PeterSmith1 @.> 发送时间: 2022年4月21日 13:22 收件人: Gumpest/YOLOv5-Multibackbone-Compression @.> 抄送: bjw1107 @.>, Author @.> 主题: Re: [Gumpest/YOLOv5-Multibackbone-Compression] SwinTrans-YOLOv5它的预权重可以是yolov5s或者yolov5m吗 (Issue #62)

zaipinai commented 2 years ago

depth_multiple: 1.33 # model depth multiple width_multiple: 1.25 # layer channel multiple 改下模型文件这俩行就能用了

请问一下，你使用这个yaml训练结果是多少阿

我的也是一般般，没有预训练权重，最后我就是加了C3TR模块，BiFPN和改了检测头用