-
I use RepVGG A2 as Backbone segmentation. After converting to engine TensorRT, float 16 keep the mIOU as the pytorch model, but when I convert to int8 TensorRT, mIOU drop ~5%. Did you try it yet?
-
Hi ppogg:
我根据你的文章《追求极致:Repvgg重参化对YOLO工业落地的实验和思考》进行了代码复现,但引入了RepVGGBlock模块后,对coco数据集训练300轮,并没有实现涨点,反而下降了。
我采用的基础版本是yolov5 5.0,主要是对yolov5s.yaml按照文章所描述的内容进行修改,并且在代码内部加入了RepVGGBlock模块…
-
首先非常感谢您的工作!在您的论文中提到在RCS-YOLO、RepVGG-CSP的消融实验,想请问一下是否是在相同的位置采用BottleneckCSPC、RCS-OSA两个模块进行实验对比?如果是我认为的这样,BottleneckCSPC仅比C3模块多一个conv1x1,那在YOLOv5中,采用RCS-OSA替换C3模块应该也可以提升模型的检测性能。请问您是否做过RCS-OSA有效性的验证实验?该模…
-
/content/yolov7
Namespace(augment=False, batch_size=16, conf_thres=0.001, data='/content/yolov7/yolov7/PV-EL-images-1/data.yaml', device='0', exist_ok=False, img_size=640, iou_thres=0.65, name='yolov…
-
### 🐛 Describe the bug
Timm_models has some models failed on accuracy check, the detail model list can be found as below table.
precision | mode | model
-- | -- | --
bfloat16 | inference | eca_b…
-
-
Feel free to check it out :smile:
https://github.com/mikel-brostrom/Yolov7_StrongSORT_OSNet
-
-
-