Open fanr520 opened 6 years ago
使用特征提取的时候,出现下标越界出错,详细说明如下: 异常发生处: 文件:FaceIdentification\include\blob.h 代码行:82 代码:return data.get()[i]; 错误类型:下标i值越界 代码当前变量值: i= 92108 count=65536 shape_=[1,1,256,256]
上一级: 文件:\FaceIdentification\src\eltwise_net.cpp 代码行:97 dst_head[offset] = (*input)[offset]
代码当前变量值: op_="BAISADDER" scale=-4.3160208e+008 upper=-4.3160208e+008 lower=-4.3160208e+008 num=1 channels=3 height=256 width=256
bn=0 bc=1 bh=1 bw=1
n=0 c=1 h=103 w=204 offset=92108
input shape=[1,1,256,256] count=65536 bias shape=[1,3,256,256] count=196608 output shape=[1,3,256,256] count=196608
上一级: 文件:\FaceIdentification\src\commonnet.cpp 代码行:174 nets[i]->Execute();
代码当前变量值: nets_ size=11 i=0
上一级: 文件:\FaceIdentification\include\recognizer.h 代码行:112 net_->Execute();
代码当前变量值: cropblob shape[1,1,256,256] count_=65536 src_img width=640 height=480 num_channels=1 points [0] 289.84335 float [1] 352.09399 float [2] 378.71692 float [3] 340.88025 float [4] 340.72888 float [5] 396.91446 float [6] 309.39282 float [7] 438.08136 float [8] 373.70685 float [9] 429.26556 float
上一级: 文件:FaceIdentification\tools\face_identification.cpp 代码行:111 recognizer->ExtractFeatureWithCrop(src_image, point_data, feats); 代码当前变量值: src_image width 640 int height 480 int num_channels 1 int point_data [0] 289.84335 float [1] 352.09399 float [2] 378.71692 float [3] 340.88025 float [4] 340.72888 float [5] 396.91446 float [6] 309.39282 float [7] 438.08136 float [8] 373.70685 float [9] 429.26556 float @seeta-wushuzhe @seetaface @jiezhang
找到原因了,是因为传入的图像能道数不对。不过软件没有做相应的提示。
我也碰到同样的问题,能不能讲下图像应该是几通道的?谢谢
通道数应该为1,要转换为通道数为1的灰度图像。
使用特征提取的时候,出现下标越界出错,详细说明如下: 异常发生处: 文件:FaceIdentification\include\blob.h 代码行:82 代码:return data.get()[i]; 错误类型:下标i值越界 代码当前变量值: i= 92108 count=65536 shape_=[1,1,256,256]
上一级: 文件:\FaceIdentification\src\eltwise_net.cpp 代码行:97 dst_head[offset] = (*input)[offset]
代码当前变量值: op_="BAISADDER" scale=-4.3160208e+008 upper=-4.3160208e+008 lower=-4.3160208e+008 num=1 channels=3 height=256 width=256
bn=0 bc=1 bh=1 bw=1
n=0 c=1 h=103 w=204 offset=92108
input shape=[1,1,256,256] count=65536 bias shape=[1,3,256,256] count=196608 output shape=[1,3,256,256] count=196608
上一级: 文件:\FaceIdentification\src\commonnet.cpp 代码行:174 nets[i]->Execute();
代码当前变量值: nets_ size=11 i=0
上一级: 文件:\FaceIdentification\include\recognizer.h 代码行:112 net_->Execute();
代码当前变量值: cropblob shape[1,1,256,256] count_=65536 src_img width=640 height=480 num_channels=1 points [0] 289.84335 float [1] 352.09399 float [2] 378.71692 float [3] 340.88025 float [4] 340.72888 float [5] 396.91446 float [6] 309.39282 float [7] 438.08136 float [8] 373.70685 float [9] 429.26556 float
上一级: 文件:FaceIdentification\tools\face_identification.cpp 代码行:111 recognizer->ExtractFeatureWithCrop(src_image, point_data, feats); 代码当前变量值: src_image width 640 int height 480 int num_channels 1 int point_data [0] 289.84335 float [1] 352.09399 float [2] 378.71692 float [3] 340.88025 float [4] 340.72888 float [5] 396.91446 float [6] 309.39282 float [7] 438.08136 float [8] 373.70685 float [9] 429.26556 float @seeta-wushuzhe @seetaface @jiezhang