search

Tencent / ncnn

ncnn is a high-performance neural network inference framework optimized for the mobile platform
Other
20.49k stars 4.17k forks source link

ncnn2table running error: Inputs is does not include all needed param #2200

Open adamantboy opened 4 years ago

adamantboy commented 4 years ago

I built ncnn using configurations as follows, but an error occurred when i ran ncnn2table, error: Inputs is does not include all needed param, pleas check.... Do you know why and how to fix it. Could you please give me some suggestions. @nihui

build configurations: cmake -DCMAKE_BUILD_TYPE=Release -DNCNN_VULKAN=ON -DNCNN_SYSTEM_GLSLANG=ON -NCNN_OPENCV=ON -NCNN_BUILD_EXAMPLES=ON ..

build output: Scanning dependencies of target ncnnoptimize [ 99%] Building CXX object tools/CMakeFiles/ncnnoptimize.dir/ncnnoptimize.cpp.o [ 99%] Built target squeezenet Scanning dependencies of target ncnn2table [ 99%] Building CXX object tools/quantize/CMakeFiles/ncnn2table.dir/ncnn2table.cpp.o [ 99%] Built target mobilenetssd [ 99%] Built target peleenetssd_seg [ 99%] Built target squeezenetssd Scanning dependencies of target ncnn2int8 [ 99%] Built target yolov3 [ 99%] Built target yolov2 [ 99%] Built target simplepose [ 99%] Building CXX object tools/quantize/CMakeFiles/ncnn2int8.dir/ncnn2int8.cpp.o [ 99%] Built target mobilenetv2ssdlite [ 99%] Built target shufflenetv2 [ 99%] Built target yolov4 [100%] Linking CXX executable ncnn2mem [100%] Built target retinaface [100%] Built target rfcn [100%] Built target fasterrcnn [100%] Built target yolact [100%] Built target ncnn2mem [100%] Linking CXX executable ncnn2int8 [100%] Built target ncnn2int8 [100%] Linking CXX executable ncnnoptimize [100%] Linking CXX executable ncnn2table [100%] Built target ncnnoptimize [100%] Built target ncnn2table

errors: --- ncnn post training quantization tool --- 13:34:46 Oct 13 2020 Inputs is does not include all needed param, pleas check... -bin b=[path to ncnn.bin file]

-help h usage ?=[print this message]

-images i=[path to calibration images folder]

-mean m=[value of mean (mean value, default is 104.0,117.0,123.0)]

-norm n=[value of normalize (scale value, default is 1.0,1.0,1.0)]

-output o=[path to output calibration table file]

-param p=[path to ncnn.param file]

-size s=[the size of input image(using the resize the original image,default is w=224,h=224)]

-swapRB c=[flag which indicates that swap first and last channels in 3-channel image is necessary]

-thread t, --4=[count of processing threads]

example: ./ncnn2table --param=squeezenet-fp32.param --bin=squeezenet-fp32.bin --images=images/ --output=squeezenet.table --mean=104.0,117.0,123.0 --norm=1.0,1.0,1.0 --size=224,224 --swapRB --thread=2

ncnn2table command: ./ncnn2table --param=yolact_resnet50_100000_550-opt.param --bin=yolact_resnet50_100000_550-opt.bin --images=/home/projects/data/COCO_dataset_2014_20200825/train2014/ --output=yolact.table --mean=123.68,116.78,103.94 --norm=0.0171232877,0.0175070028,0.0174276751 --size=550,550 --swapRB --thread=12

operating system: Unbuntu16.04

adamantboy commented 4 years ago

I am sure that the inputs param are all included.

zhongqianli commented 3 years ago

I built ncnn using configurations as follows, but an error occurred when i ran ncnn2table, error: Inputs is does not include all needed param, pleas check.... Do you know why and how to fix it. Could you please give me some suggestions. @nihui

build configurations: cmake -DCMAKE_BUILD_TYPE=Release -DNCNN_VULKAN=ON -DNCNN_SYSTEM_GLSLANG=ON -NCNN_OPENCV=ON -NCNN_BUILD_EXAMPLES=ON ..

build output: Scanning dependencies of target ncnnoptimize [ 99%] Building CXX object tools/CMakeFiles/ncnnoptimize.dir/ncnnoptimize.cpp.o [ 99%] Built target squeezenet Scanning dependencies of target ncnn2table [ 99%] Building CXX object tools/quantize/CMakeFiles/ncnn2table.dir/ncnn2table.cpp.o [ 99%] Built target mobilenetssd [ 99%] Built target peleenetssd_seg [ 99%] Built target squeezenetssd Scanning dependencies of target ncnn2int8 [ 99%] Built target yolov3 [ 99%] Built target yolov2 [ 99%] Built target simplepose [ 99%] Building CXX object tools/quantize/CMakeFiles/ncnn2int8.dir/ncnn2int8.cpp.o [ 99%] Built target mobilenetv2ssdlite [ 99%] Built target shufflenetv2 [ 99%] Built target yolov4 [100%] Linking CXX executable ncnn2mem [100%] Built target retinaface [100%] Built target rfcn [100%] Built target fasterrcnn [100%] Built target yolact [100%] Built target ncnn2mem [100%] Linking CXX executable ncnn2int8 [100%] Built target ncnn2int8 [100%] Linking CXX executable ncnnoptimize [100%] Linking CXX executable ncnn2table [100%] Built target ncnnoptimize [100%] Built target ncnn2table

errors: --- ncnn post training quantization tool --- 13:34:46 Oct 13 2020 Inputs is does not include all needed param, pleas check... -bin b=[path to ncnn.bin file]

-help h usage ?=[print this message]

-images i=[path to calibration images folder]

-mean m=[value of mean (mean value, default is 104.0,117.0,123.0)]

-norm n=[value of normalize (scale value, default is 1.0,1.0,1.0)]

-output o=[path to output calibration table file]

-param p=[path to ncnn.param file]

-size s=[the size of input image(using the resize the original image,default is w=224,h=224)]

-swapRB c=[flag which indicates that swap first and last channels in 3-channel image is necessary]

-thread t, --4=[count of processing threads]

example: ./ncnn2table --param=squeezenet-fp32.param --bin=squeezenet-fp32.bin --images=images/ --output=squeezenet.table --mean=104.0,117.0,123.0 --norm=1.0,1.0,1.0 --size=224,224 --swapRB --thread=2

ncnn2table command: ./ncnn2table --param=yolact_resnet50_100000_550-opt.param --bin=yolact_resnet50_100000_550-opt.bin --images=/home/projects/data/COCO_dataset_2014_20200825/train2014/ --output=yolact.table --mean=123.68,116.78,103.94 --norm=0.0171232877,0.0175070028,0.0174276751 --size=550,550 --swapRB --thread=12

operating system: Unbuntu16.04

I built ncnn using configurations as follows, but an error occurred when i ran ncnn2table, error: Inputs is does not include all needed param, pleas check.... Do you know why and how to fix it. Could you please give me some suggestions. @nihui

build configurations: cmake -DCMAKE_BUILD_TYPE=Release -DNCNN_VULKAN=ON -DNCNN_SYSTEM_GLSLANG=ON -NCNN_OPENCV=ON -NCNN_BUILD_EXAMPLES=ON ..

build output: Scanning dependencies of target ncnnoptimize [ 99%] Building CXX object tools/CMakeFiles/ncnnoptimize.dir/ncnnoptimize.cpp.o [ 99%] Built target squeezenet Scanning dependencies of target ncnn2table [ 99%] Building CXX object tools/quantize/CMakeFiles/ncnn2table.dir/ncnn2table.cpp.o [ 99%] Built target mobilenetssd [ 99%] Built target peleenetssd_seg [ 99%] Built target squeezenetssd Scanning dependencies of target ncnn2int8 [ 99%] Built target yolov3 [ 99%] Built target yolov2 [ 99%] Built target simplepose [ 99%] Building CXX object tools/quantize/CMakeFiles/ncnn2int8.dir/ncnn2int8.cpp.o [ 99%] Built target mobilenetv2ssdlite [ 99%] Built target shufflenetv2 [ 99%] Built target yolov4 [100%] Linking CXX executable ncnn2mem [100%] Built target retinaface [100%] Built target rfcn [100%] Built target fasterrcnn [100%] Built target yolact [100%] Built target ncnn2mem [100%] Linking CXX executable ncnn2int8 [100%] Built target ncnn2int8 [100%] Linking CXX executable ncnnoptimize [100%] Linking CXX executable ncnn2table [100%] Built target ncnnoptimize [100%] Built target ncnn2table

errors: --- ncnn post training quantization tool --- 13:34:46 Oct 13 2020 Inputs is does not include all needed param, pleas check... -bin b=[path to ncnn.bin file]

-help h usage ?=[print this message]

-images i=[path to calibration images folder]

-mean m=[value of mean (mean value, default is 104.0,117.0,123.0)]

-norm n=[value of normalize (scale value, default is 1.0,1.0,1.0)]

-output o=[path to output calibration table file]

-param p=[path to ncnn.param file]

-size s=[the size of input image(using the resize the original image,default is w=224,h=224)]

-swapRB c=[flag which indicates that swap first and last channels in 3-channel image is necessary]

-thread t, --4=[count of processing threads]

example: ./ncnn2table --param=squeezenet-fp32.param --bin=squeezenet-fp32.bin --images=images/ --output=squeezenet.table --mean=104.0,117.0,123.0 --norm=1.0,1.0,1.0 --size=224,224 --swapRB --thread=2

ncnn2table command: ./ncnn2table --param=yolact_resnet50_100000_550-opt.param --bin=yolact_resnet50_100000_550-opt.bin --images=/home/projects/data/COCO_dataset_2014_20200825/train2014/ --output=yolact.table --mean=123.68,116.78,103.94 --norm=0.0171232877,0.0175070028,0.0174276751 --size=550,550 --swapRB --thread=12

operating system: Unbuntu16.04

Use old version of ncnn2table.cpp is OK. Argument parser has bug. So you can try to git checkout to old version ncnn.