#921 #994 - fix handling in DigitalUebergangsbereichVorgaenger, reduced DigitalUebergangsbereichVorgaenger =0.7

added serial-debug in postprocessing (not enabled by default)
added test-case for #994
fixed checkDigitConsistency (pre value before analog lost)
added debug to proof need of double for NUMBERS.Value (see testcase 32289.4198)
This commit is contained in:
Frank Haverland
2022-09-11 22:44:35 +02:00
parent dd70aa8969
commit d7c7537aa9
4 changed files with 136 additions and 20 deletions

View File

@@ -5,7 +5,8 @@
#include <ClassFlowMakeImage.h>
void setUpClassFlowPostprocessing(void);
string process_doFlow(std::vector<float> analog, std::vector<float> digits, t_CNNType digType = Digital100);
string process_doFlow(std::vector<float> analog, std::vector<float> digits, t_CNNType digType = Digital100,
bool checkConsistency=false, bool extendedResolution=false, int decimal_shift=0);
ClassFlowCNNGeneral* _analog;
ClassFlowCNNGeneral* _digit;
@@ -43,6 +44,7 @@ void test_doFlow() {
std::vector<float> digits = { 1.2, 6.7};
std::vector<float> analogs = { 9.5, 8.4};
const char* expected = "16.98";
const char* expected_extended = "16.984";
std::string result = process_doFlow(analogs, digits);
TEST_ASSERT_EQUAL_STRING(expected, result.c_str());
@@ -195,10 +197,70 @@ void test_doFlow() {
// https://github.com/jomjol/AI-on-the-edge-device/issues/921#issuecomment-1242730397
digits = { 3.0, 2.0, 2.0, 8.0, 9.0, 4.0, 1.7, 9.8}; // falscher Wert 32290.420
analogs = { };
expected = "32289420";
result = process_doFlow(analogs, digits);
expected = "32289.420";
expected_extended= "32289.4199";
// FALSCH! wegen ungenügender Präzision von NUMBERS->Value
// expected_extended= "32289.4198";
// extendResolution=false, checkConsistency=false
result = process_doFlow(analogs, digits, Digital100, false, false, -3);
TEST_ASSERT_EQUAL_STRING(expected, result.c_str());
// extendResolution=true
result = process_doFlow(analogs, digits, Digital100, false, true, -3);
TEST_ASSERT_EQUAL_STRING(expected_extended, result.c_str());
// checkConsistency=true und extendResolution=true
result = process_doFlow(analogs, digits, Digital100, false, true, -3);
TEST_ASSERT_EQUAL_STRING(expected_extended, result.c_str());
// Fehler Rolling (2022-09-10)
// not documented as issue
digits = { 0.0, 0.0, 7.9, 3.8}; // 84.99401 als falsches Ergebnis
analogs = { 0.0, 9.4, 4.1, 0.1};
expected = "83.9940";
expected_extended= "83.99401";
// checkConsistency=false
result = process_doFlow(analogs, digits, Digital100, false);
TEST_ASSERT_EQUAL_STRING(expected, result.c_str());
// checkConsistency=true
result = process_doFlow(analogs, digits, Digital100, true);
TEST_ASSERT_EQUAL_STRING(expected, result.c_str());
// extendResolution=true
result = process_doFlow(analogs, digits, Digital100, false, true);
TEST_ASSERT_EQUAL_STRING(expected_extended, result.c_str());
// checkConsistency=true und extendResolution=true
result = process_doFlow(analogs, digits, Digital100, false, true);
TEST_ASSERT_EQUAL_STRING(expected_extended, result.c_str());
// Fehler Rolling (2022-09-10)
// https://github.com/jomjol/AI-on-the-edge-device/issues/994#issue-1368570945
digits = { 0.0, 0.0, 1.0, 2.0, 2.8, 1.9, 2.8, 5.6}; // 123245.6 als falsches Ergebnis
analogs = { };
expected = "123236";
expected_extended= "123235.6";
// checkConsistency=true
result = process_doFlow(analogs, digits, Digital100, false, false);
TEST_ASSERT_EQUAL_STRING(expected, result.c_str());
// checkConsistency=true
result = process_doFlow(analogs, digits, Digital100, true, false);
TEST_ASSERT_EQUAL_STRING(expected, result.c_str());
// extendResolution=true
result = process_doFlow(analogs, digits, Digital100, false, true);
TEST_ASSERT_EQUAL_STRING(expected_extended, result.c_str());
// checkConsistency=true und extendResolution=true
result = process_doFlow(analogs, digits, Digital100, false, true);
TEST_ASSERT_EQUAL_STRING(expected_extended, result.c_str());
}
@@ -221,7 +283,8 @@ void setUpClassFlowPostprocessing(t_CNNType digType, t_CNNType anaType)
}
std::string process_doFlow(std::vector<float> analog, std::vector<float> digits, t_CNNType digType) {
std::string process_doFlow(std::vector<float> analog, std::vector<float> digits, t_CNNType digType,
bool checkConsistency, bool extendedResolution, int decimal_shift) {
// setup the classundertest
setUpClassFlowPostprocessing(digType, Analogue100);
@@ -231,7 +294,6 @@ std::string process_doFlow(std::vector<float> analog, std::vector<float> digits,
if (digits.size()>0) {
general* gen_digit = _digit->GetGENERAL("default", true);
gen_digit->ROI.clear();
for (int i = 0; i<digits.size(); i++) {
roi* digitROI = new roi();
string name = "digit_" + std::to_string(i);
@@ -258,6 +320,30 @@ std::string process_doFlow(std::vector<float> analog, std::vector<float> digits,
printf("Setup ROIs completed.\n");
undertestPost->InitNUMBERS();
if (checkConsistency) {
printf("checkConsistency=true\n");
std::vector<NumberPost*>* NUMBERS = undertestPost->GetNumbers();
for (int _n = 0; _n < (*NUMBERS).size(); ++_n) {
printf("Set checkConsistency on number: %d\n", _n);
(*NUMBERS)[_n]->checkDigitIncreaseConsistency = true;
}
}
if (extendedResolution ) {
std::vector<NumberPost*>* NUMBERS = undertestPost->GetNumbers();
for (int _n = 0; _n < (*NUMBERS).size(); ++_n) {
printf("Set extendedResolution on number: %d\n", _n);
(*NUMBERS)[_n]->isExtendedResolution = true;
}
}
if (decimal_shift!=0) {
std::vector<NumberPost*>* NUMBERS = undertestPost->GetNumbers();
for (int _n = 0; _n < (*NUMBERS).size(); ++_n) {
printf("Set decimalshif on number: %d to %d\n", _n, decimal_shift);
(*NUMBERS)[_n]->DecimalShift = decimal_shift;
(*NUMBERS)[_n]->DecimalShiftInitial = decimal_shift;
}
}
string time;
// run test