AnsweredAssumed Answered

้how to make file  BLACKFIN HOG SVM DETECTOR   56 X 112 .DAT

Question asked by THEZOOZ on Nov 8, 2015

I study   BLACKFIN HOG SVM DETECTOR .

I   follow  in datasheet  step byp step by  KT-2383_HOGSVM_UsersGuide.pdf

 

adi_hogsvm_dump_features.exe -i "INRIAPerson/96X160H96/Train/pos/" -o train_pos_feature_64x128.dat -n 0 -w 64 -h 128 -v 2 -kt 2 -ci 1 -nw 0 -xs 0 -ys 0 -sr 1.200000 -ss  1.000000 -se 0.000000 -a 0 -fl 0 -ol train_positive_vectors_64x128.txt

 

adi_hogsvm_dump_features.exe -i "INRIAPerson/Train/neg/" -o train_neg_feature_64x128.dat -n 0 -w 64 -h 128 -v 2 -kt 2 -ci 0 -nw 10 -xs 0 -ys 0 -sr 1.200000 -ss  1.000000 -se 0.000000 -a 0 -fl 0 -ol train_negative_vectors_64x128.txt

 

 

adi_hogsvm_train.exe -p train_pos_feature_64x128.dat -n train_neg_feature_64x128.dat -o primary_64x128.xml -nn 0 -np 0 -c 0.010000 -tm 0 -ti 1000  -tc 2 -te 0.100000 -v 2  -pw 1.000000 -ef training_error_64x128.txt

 

 

adi_hogsvm_dump_features.exe -i "INRIAPerson/Train/neg/" -o hardneg_feature_64x128.dat -n 0 -w 64 -h 128 -v 2 -kt 2 -ci 0 -nw 0 -xs 8 -ys 8 -sr 1.200000 -ss  1.000000 -se 4.000000 -a 0 -fl 0 -ol hard_negative_vectors_64x128.txt -t primary_64x128.xml

 

adi_hogsvm_train.exe  -p train_pos_feature_64x128.dat  -n train_neg_feature_64x128.dat -o hard_primary_64x128.xml -nn 0 -np 0  -c 0.010000 -tm 0 -ti 6000  -tc 2 -te 0.001000 -v 2 -pw 3.000000 -ef training_error_64x128.txt  -hn hardneg_feature_64x128.dat

 

adi_hogsvm_convert.exe -kt 2 -t hard_primary_64x128.xml -tf inria_model_64x128.dat -w 64 -h 128

 

  1. mode 64X128.dat  size 7KB   is    same  example in  C: Analog Devices\SoftwareModules\HOGSVMDetector-BF-Rel1.1.0\demo\utils ß  it is ok

 

I try  make  file demo_model_56X112.dat    in  C: Analog Devices\SoftwareModules\HOGSVMDetector-BF-Rel1.1.0\demo\utils is 10 KB    but   I make  7 kb  <-  not work

 

adi_hogsvm_dump_features.exe -i  "INRIAPerson/96X160H96/Train/pos/" -o train_pos_feature_56x112.dat -n 0 -w 56 -h 112 -v 2 -kt 2 -ci 1 -nw 0  -xs 0 -ys 0 -sr 1.200000 -ss  1.000000 -se 0.000000 -a 0 -fl 0 -ol train_positive_vectors_56x112.txt

 

adi_hogsvm_dump_features.exe -i  "INRIAPerson/Train/neg/" -o train_neg_feature_56x112.dat -n 0 -w 56 -h 112 -v 2 -kt 2 -ci 0 -nw 10 -xs 0 -ys 0 -sr 1.200000 -ss  1.000000 -se 0.000000 -a 0 -fl 0 -ol train_negative_vectors_56x112.txt

 

adi_hogsvm_train.exe -p train_pos_feature_56x112.dat -n train_neg_feature_56x112.dat -o primary_56x112.xml -nn 0 -np 0 -c 0.010000 -tm 0 -ti 1000  -tc 2 -te 0.100000 -v 2  -pw 1.000000 -ef training_error_56x112.txt

 

adi_hogsvm_dump_features.exe -i  "INRIAPerson/Train/neg/" -o hardneg_feature_56x112.dat -n 0 -w 56 -h 112 -v 2 -kt 2 -ci 0 -nw 0 -xs 8 -ys 8 -sr 1.400000 -ss  1.000000 -se 4.000000 -a 0 -fl 0 -ol hard_negative_vectors_56x112.txt -t primary_56x112.xml

 

adi_hogsvm_train.exe  -p train_pos_feature_56x112.dat  -n train_neg_feature_56x112.dat -o hard_primary_56x112.xml -nn 0 -np 0  -c 0.010000 -tm 0 -ti 6000  -tc 2 -te 0.001000 -v 2 -pw 3.000000 -ef training_error_56x112.txt  -hn hardneg_feature_56x112.dat

adi_hogsvm_convert.exe -kt 2 -t hard_primary_56x112.xml -tf inria_model_56x112.dat -w 56 -h 112

Outcomes