From 66adeedf5da99bd5bfa103791fc33afd03abdd18 Mon Sep 17 00:00:00 2001
From: Micooz <Micooz@hotmail.com>
Date: Tue, 11 Aug 2015 20:06:20 +0800
Subject: [PATCH] Finished ANN training for opencv3.0.
---
src/train/ann_train.cpp | 192 +++++++++++++++++++++++-----------------
1 file changed, 113 insertions(+), 79 deletions(-)
diff --git a/src/train/ann_train.cpp b/src/train/ann_train.cpp
index 19ad9ff..bdbeac6 100644
--- a/src/train/ann_train.cpp
+++ b/src/train/ann_train.cpp
@@ -1,88 +1,122 @@
-#include "easypr/train/ann_train.h"
+#include "easypr/train/ann_train.h"
+#include "easypr/core/core_func.h"
+#include "easypr/core/chars_identify.h"
+#include "easypr/util/util.h"
#include "easypr/config.h"
-
-namespace easypr{
-
- AnnTrain::AnnTrain(const char* chars_folder, const char* zhchars_folder, const char* xml)
- :chars_folder_(chars_folder), zhchars_folder_(zhchars_folder), ann_xml_(xml){
-
- }
-
- void AnnTrain::train(const int & neurons /* = 40 */){
- this->getTrainData();
-
- cv::Mat layers = {
- train_data_->getNSamples(),// the input layer
- neurons, // the neurons
- sizeof(kChinese) + sizeof(kCharacters) // the output layer
- };
- ann_->setLayerSizes(layers);
- ann_->setTrainMethod(cv::ml::ANN_MLP::TrainingMethods::BACKPROP);
- ann_->setBackpropWeightScale(0.1);
- ann_->setBackpropMomentumScale(0.1);
-
- std::cout << "Training ANN model, please wait..." << std::endl;
- long start = utils::getTimestamp();
- ann_->train(train_data_);
- long end = utils::getTimestamp();
- std::cout << "Training done. Elapse: " << (end - start) / 1000 << std::endl;
-
- ann_->save(ann_xml_);
- std::cout << "Your ANN Model was saved to " << ann_xml_ << std::endl;
- }
-
- void AnnTrain::getTrainData(){
- assert(chars_folder_);
- assert(zhchars_folder_);
- // create new
- cv::Mat samples;
- cv::Mat responses;
-
- std::cout << "Collecting chars in " << chars_folder_ << std::endl;
-
- for (auto i = 0; i < sizeof(kCharacters); ++i){
- char c = kCharacters[i];
- char sub_folder[512] = { 0 };
-
- sprintf(sub_folder, "%s/%c", chars_folder_, c);
- std::cout << " >> Featuring characters " << c << " in " << sub_folder << std::endl;
-
- auto chars_files = utils::getFiles(sub_folder);
- for (auto file : chars_files){
- auto img = cv::imread(file);
- auto fps = features(img, kPredictSize);
-
- samples.push_back(fps);
- responses.push_back(i);
- }
- }
- std::cout << "Collecting zh-chars in " << zhchars_folder_ << std::endl;
+namespace easypr {
+
+AnnTrain::AnnTrain(const char* chars_folder,
+ const char* xml)
+ : chars_folder_(chars_folder),
+ ann_xml_(xml) {
+ ann_ = cv::ml::ANN_MLP::create();
+}
+
+void AnnTrain::train(const int& neurons /* = 40 */) {
+ cv::Mat layers(1, 3, CV_32SC1);
+ layers.at<int>(0) = 120; // the input layer
+ layers.at<int>(1) = neurons; // the neurons
+ layers.at<int>(2) = kCharsTotalNumber; // the output layer
+
+ ann_->setLayerSizes(layers);
+ ann_->setActivationFunction(cv::ml::ANN_MLP::SIGMOID_SYM, 1, 1);
+ ann_->setTrainMethod(cv::ml::ANN_MLP::TrainingMethods::BACKPROP);
+ ann_->setBackpropWeightScale(0.1);
+ ann_->setBackpropMomentumScale(0.1);
+
+ auto traindata = train_data();
+ std::cout << "Training ANN model, please wait..." << std::endl;
+ long start = utils::getTimestamp();
+ ann_->train(traindata);
+ long end = utils::getTimestamp();
+ std::cout << "Training done. Time elapse: " << (end - start) << "ms" << std::endl;
+
+ ann_->save(ann_xml_);
+ std::cout << "Your ANN Model was saved to " << ann_xml_ << std::endl;
+}
+
+cv::Ptr<cv::ml::TrainData> AnnTrain::train_data() {
+ assert(chars_folder_);
- for (auto i = 0; i < sizeof(kChinese); ++i){
- const char *zhc = kChinese[i];
- char sub_folder[512] = { 0 };
+ cv::Mat samples;
+ std::vector<int> labels;
- sprintf(sub_folder, "%s/%s", zhchars_folder_, zhc);
- std::cout << " >> Featuring zh-characters " << zhc << " in " << sub_folder << std::endl;
+ std::cout << "Collecting chars in " << chars_folder_ << std::endl;
- auto chars_files = utils::getFiles(sub_folder);
- for (auto file : chars_files){
- auto img = cv::imread(file);
- auto fps = features(img, kPredictSize);
+ for (int i = 0; i < kCharsTotalNumber; ++i) {
+ auto char_key = kChars[i];
+ char sub_folder[512] = {0};
- samples.push_back(fps);
- responses.push_back(i + sizeof(kCharacters));
+ sprintf(sub_folder, "%s/%s", chars_folder_, char_key);
+ std::cout << " >> Featuring characters " << char_key << " in " << sub_folder << std::endl;
+
+ auto chars_files = utils::getFiles(sub_folder);
+ for (auto file : chars_files) {
+ auto img = cv::imread(file);
+ auto fps = features(img, kPredictSize);
+
+ samples.push_back(fps);
+ labels.push_back(i);
+ }
+ }
+
+ cv::Mat samples_;
+ samples.convertTo(samples_, CV_32F);
+ cv::Mat train_classes = cv::Mat::zeros((int) labels.size(), kCharsTotalNumber,
+ CV_32F);
+
+ for (int i = 0; i < train_classes.rows; ++i) {
+ train_classes.at<float>(i, labels[i]) = 1.f;
+ }
+
+ return cv::ml::TrainData::create(samples_,
+ cv::ml::SampleTypes::ROW_SAMPLE,
+ train_classes);
+}
+
+void AnnTrain::test() {
+ assert(chars_folder_);
+
+ for (int i = 0; i < kCharsTotalNumber; ++i) {
+ auto char_key = kChars[i];
+ char sub_folder[512] = {0};
+
+ sprintf(sub_folder, "%s/%s", chars_folder_, char_key);
+ fprintf(stdout, " >> Testing characters %s in %s \n",
+ char_key, sub_folder);
+
+ auto chars_files = utils::getFiles(sub_folder);
+ int corrects = 0, sum = 0;
+ std::vector<std::string> error_files;
+
+ for (auto file : chars_files) {
+ auto img = cv::imread(file);
+ std::pair<std::string, std::string> ch = CharsIdentify::instance()->identify(img);
+ if (ch.first == char_key) {
+ ++corrects;
+ } else {
+ error_files.push_back(utils::getFileName(file));
}
+ ++sum;
}
- //
- cv::Mat samples_row;
- cv::Mat responses_row;
-
- samples.convertTo(samples_row, CV_32FC1);
- responses.convertTo(responses_row, CV_32FC1);
+ fprintf(stdout, " >> [sum: %d, correct: %d, rate: %.2f]\n",
+ sum, corrects, (float)corrects / (sum == 0 ? 1 : sum));
+ std::string error_string;
+ auto end = error_files.end();
+ if (error_files.size() >= 10) {
+ end -= error_files.size() * (1 - 0.1);
+ }
+ for(auto i = error_files.begin(); i != end; ++i) {
+ error_string.append(*i);
+ if (i != end - 1) {
+ error_string.append(", ");
+ } else {
+ error_string.append(" ...");
+ }
+ }
+ fprintf(stdout, " >> [%s]\n", error_string.c_str());
+ }
+}
- train_data_ = cv::ml::TrainData::create(samples_row, ml::SampleTypes::ROW_SAMPLE, responses_row);
- }
-
-}
+}