yolov3转换 Upsample Layer出错

ChenYingpeng / darknet2caffe

Convert darknet weights to caffemodel

183 stars 88 forks source link

yolov3转换 Upsample Layer出错 #11

Open zgq91 opened 4 years ago

zgq91 commented 4 years ago

hi: 感谢开源我在转成darknet官方的yolov3.cfg yolov3.weights时出错:Check failed: ExactNumBottomBlobs() == bottom.size() (2 vs. 1) Upsample Layer takes 2 bottom blob(s) as input.环境windows10 python3.6

ChenYingpeng commented 4 years ago

Please check your upsample layer which botton blobs size must be 1. Such as `

layer { bottom: "layer85-conv" top: "layer86-upsample" name: "layer86-upsample" type: "Upsample" upsample_param { scale: 2 } } `

yulinhuyang commented 4 years ago

layer { bottom: "layer118-conv" top: "layer119-upsample" name: "layer119-upsample" type: "Upsample" upsample_param { scale: 2 } } 这是我的转换结果，在一些嵌入式AI环境仿真也报同样错误，没看明白为啥。

yulinhuyang commented 4 years ago

看到网上又用deconv替换upsample的。不知道能否解决

ChenYingpeng commented 4 years ago

Please post more details about your error log.