Part Number: PROCESSOR-SDK-DRA8X-TDA4X
Tool/software: Code Composer Studio
HI:
I am using psdk_rtos_auto_j7_06_02_00_21/tidl_j7_01_01_00_10 tidl_model_import my CNN and i dump the data in psdk_rtos_auto_j7_06_02_00_21/tidl_j7_01_01_00_10/ti_dl/test/trace .
i have set inDataNorm = 1 inMean = 128.0 128.0 128.0 inScale = 0.00390625 0.00390625 0.00390625 but it is invalid.
i have check TIDL_DataLayer output Still the pixel value of the image.
How to make normalization effective ？
follow i give out my tidl_import and tidl_infer parameter.
///////////////////////////////////////tidl_import parameter
modelType = 0
inputNetFile = "../../test/testvecs/models/public/caffe/test_simple1/test_simple1.prototxt"
inputParamsFile = "../../test/testvecs/models/public/caffe/test_simple1/test_simple1.caffemodel"
outputNetFile = "../../test/testvecs/config/tidl_models/caffe/test/tidl_net_test.bin"
outputParamsFile = "../../test/testvecs/config/tidl_models/caffe/test/tidl_io_test_"
inDataNorm = 1
inMean = 128.0 128.0 128.0
inScale = 0.00390625 0.00390625 0.00390625
inFileFormat = 0
inDataFormat = 1
inData = "../../test/testvecs/input/test.jpg"
postProcType = 0
perfSimConfig = ../../test/testvecs/config/import/perfsim_base.cfg
inWidth = 128
inHeight = 128
inNumChannels = 3
foldMaxPoolInConv2D = 0
/////////////////////////////////////////////tidl_infer parameter.
numFrames = 1
netBinFile = "testvecs/config/tidl_models/caffe/test/tidl_net_test.bin"
ioConfigFile = "testvecs/config/tidl_models/caffe/test/tidl_io_test_1.bin"
outData = "testvecs/output/test.bin"
inFileFormat = 0
#inDataFormat = 1
inData = "testvecs/input/test.jpg"
debugTraceLevel = 3
writeTraceLevel = 3
enableLayerPerfTraces = 1
flowCtrl = 4
#reservedCtrl = 2
Thanks!
Shuai
