mark/small-image-descriptors
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

First two descriptors migrated

Browse Source
This commit is contained in:
Mark Hoekveen
2019-06-04 19:44:49 +02:00
parent 375626e6f6
commit 9bde653a2a
4 changed files with 76 additions and 81 deletions
Download Patch File Download Diff File Expand all files Collapse all files
Descriptor.h
+43 -43
View File
@@ -16,11 +16,11 @@ namespace fs = boost::filesystem;
//but should be overloaded by a child class
class Descriptor {
public:
virtual bitset<64> feature(Image* img) = 0;
virtual uint64_t feature(Image* img) = 0;
//Interpret the bitsets as integers and calculate their distance.
virtual unsigned long long distance(bitset<64> a, bitset<64> b) {
return abs(((long long)a.to_ullong() - (long long)b.to_ullong()));
virtual uint64_t distance(uint64_t a, uint64_t b) {
return abs((long long)a - (long long)b);
};
virtual string ToString() = 0;
virtual fs::path ToPath() {
@@ -28,7 +28,7 @@ public:
}
//should return true if a > b
virtual bool bigger(bitset<64> a, bitset<64> b) {
virtual bool bigger(uint64_t a, uint64_t b) {
return true;
}
@@ -49,7 +49,7 @@ public:
// Pixels view(*img);
// int chan = (int)img->channels(); //should be 1, just intensity
// const Quantum *p = view.getConst(0, 0, n, m); //entire image
// bitset<64> retval(0);
// uint64_t retval(0);
// int N = n*m;
// Image* newimg = new Image(Geometry(n, m), Color("white"));
// newimg->type(GrayscaleType);
@@ -92,14 +92,14 @@ public:
//with 8 bits per color.
class averageColor : public Descriptor {
public:
bitset<64> feature(Image* img) {
uint64_t feature(Image* img) {
ssize_t n = img->columns();
ssize_t m = img->rows();
Pixels view(*img);
const Quantum *p = view.getConst(0, 0, n, m); //entire image
unsigned long long red = 0, green = 0, blue = 0;
float numpixels = (float)(n*m);
if (img->channels() < 3) return bitset<64>(0);
if (img->channels() < 3) return 0;
for (ssize_t j = 0; j < m; j++) {
for (ssize_t i = 0; i < n; i++) {
//LTR scanning
@@ -114,36 +114,36 @@ public:
int g = ((green/numpixels) / QuantumRange) * 255;
int b = ((blue/numpixels) / QuantumRange) * 255;
//now store all this into a bitset.. assuming 8 bits per color, if everything went correctly.. which we should check.. probably..
bitset<64> retval(r);
uint64_t retval = r;
retval <<= 8;
retval |= bitset<64>(g);
retval += g;
retval <<= 8;
retval |= bitset<64>(b);
retval += b;
return retval;
}
unsigned long long distance(bitset<64> a, bitset<64> b) {
uint64_t distance(uint64_t a, uint64_t b) {
//do some bit shifting magic to get ints back
int red_a = (a >> 16).to_ulong();
int green_a = (a << (64-16) >> (64-8)).to_ulong();
int blue_a = (a << (64-8) >> (64-8)).to_ulong();
int red_b = (b >> 16).to_ulong();
int green_b = (b << (64 - 16) >> (64 - 8)).to_ulong();
int blue_b = (b << (64 - 8) >> (64 - 8)).to_ulong();
int red_a = (a >> 16);
int green_a = (a << (64-16) >> (64-8));
int blue_a = (a << (64-8) >> (64-8));
int red_b = (b >> 16);
int green_b = (b << (64 - 16) >> (64 - 8));
int blue_b = (b << (64 - 8) >> (64 - 8));
return abs(red_a - red_b) + abs(green_a - green_b) + abs(blue_a - blue_b);
}
string ToString() {
return "averageColor";
}
};
/*
//Counts the total number of colors
//and stores it in bit representation.
//Upper limit 16777216
//Only uses least significant 24 bits
class numColors : public Descriptor {
public:
bitset<64> feature(Image* img) {
return bitset<64>(img->totalColors()-1);
uint64_t feature(Image* img) {
return uint64_t(img->totalColors()-1);
}
string ToString() {
return "numColors";
@@ -155,7 +155,7 @@ public:
//higher than median and 0 otherwise.
class median : public Descriptor {
public:
bitset<64> feature(Image* img) {
uint64_t feature(Image* img) {
Image* local_image = new Image(*img);
local_image->type(GrayscaleType);
local_image->sample(Geometry("8x8!")); //64 pixels
@@ -181,13 +181,13 @@ public:
sort(intensities, intensities + numpixels);
//We know where the median is going to be:
intensity = (float)(intensities[31] + intensities[32]) / 2;
bitset<64> retval(0);
uint64_t retval(0);
const Quantum *q = view.getConst(0, 0, n, m); //entire image
for (ssize_t j = 0; j < m; j++) {
for (ssize_t i = 0; i < n; i++) {
retval <<= 1;
if (q[0] > intensity)
retval |= bitset<64>(1);
retval |= uint64_t(1);
q = q + local_image->channels();
}
}
@@ -197,7 +197,7 @@ public:
}
//biterror
unsigned long long distance(bitset<64> a, bitset<64> b) {
unsigned long long distance(uint64_t a, uint64_t b) {
int biterror = 0;
for (int i = 0; i < 64; i++) {
if (a[i] != b[i]) biterror++;
@@ -214,7 +214,7 @@ public:
//sobel convolution applied in both directions
class sobel : public Descriptor {
public:
bitset<64> feature(Image* img) {
uint64_t feature(Image* img) {
Image* local_image = new Image(*img);
//TODO: Some preprocessing for this?
//img->reduceNoise(4.0);
@@ -242,7 +242,7 @@ public:
}
int scaledIntensity = ((float)(intensity / (n*m)) / QuantumRange) * 255;
delete local_image;
return bitset<64>(scaledIntensity);
return uint64_t(scaledIntensity);
}
string ToString() {
return "sobel";
@@ -258,7 +258,7 @@ private:
//0 otherwise
class pearson : public Descriptor {
public:
bitset<64> feature(Image* img) {
uint64_t feature(Image* img) {
Image* local_image = new Image(*img);
local_image->type(GrayscaleType);
local_image->sample(Geometry("8x8!")); //64 pixels
@@ -267,14 +267,14 @@ public:
Pixels view(*local_image);
int chan = (int)local_image->channels();
const Quantum *p = view.getConst(0, 0, n, m); //entire image
bitset<64> retval(0);
uint64_t retval(0);
int prev = p[0]; //so we always start with a 0
for (ssize_t j = 0; j < m; j++) {
for (ssize_t i = 0; i < n; i++) {
//LTR scanning
//NOTE: Scanning order might be quite relevant for this descriptor
int cur = p[0];
if(prev < cur) retval |= bitset<64>(1); //XOR
if(prev < cur) retval |= uint64_t(1); //XOR
prev = cur;
p = p + chan;
@@ -284,7 +284,7 @@ public:
delete local_image;
return retval;
}
unsigned long long distance(bitset<64> a, bitset<64> b) {
unsigned long long distance(uint64_t a, uint64_t b) {
int biterror = 0;
for (int i = 0; i < 64; i++) {
if (a[i] != b[i]) biterror++;
@@ -301,7 +301,7 @@ public:
//neighbourhood and stores it in a LTR map
class threshold : public Descriptor {
public:
bitset<64> feature(Image* img) {
uint64_t feature(Image* img) {
Image* local_image = new Image(*img);
local_image->type(GrayscaleType);
local_image->sample(Geometry("8x8!")); //64 pixels
@@ -311,11 +311,11 @@ public:
Pixels view(*local_image);
int chan = (int)local_image->channels();
const Quantum *p = view.getConst(0, 0, n, m); //entire image
bitset<64> retval(0);
uint64_t retval(0);
for (ssize_t j = 0; j < m; j++) {
for (ssize_t i = 0; i < n; i++) {
if(p[0] != 0)
retval |= bitset<64>(1); //XOR
retval |= uint64_t(1); //XOR
retval <<= 1;
p = p + chan;
@@ -324,7 +324,7 @@ public:
delete local_image;
return retval;
}
unsigned long long distance(bitset<64> a, bitset<64> b) {
unsigned long long distance(uint64_t a, uint64_t b) {
int biterror = 0;
for (int i = 0; i < 64; i++) {
if (a[i] != b[i]) biterror++;
@@ -382,15 +382,15 @@ public:
//neighbourhood and stores it in a LTR map
class DCTPearson : public DCTDescriptor {
public:
bitset<64> feature(Image* img) {
uint64_t feature(Image* img) {
Image* local_image = new Image(*img);
double* converted = CalculateDCT(local_image);
double prev = 0;
bitset<64> retval(0);
uint64_t retval(0);
for (int i = 0; i < 64; i++) {
if (converted[i] > prev) {
retval |= bitset<64>(1); //XOR
retval |= uint64_t(1); //XOR
}
retval <<= 1;
prev = converted[i];
@@ -398,7 +398,7 @@ public:
delete local_image;
return retval;
}
unsigned long long distance(bitset<64> a, bitset<64> b) {
unsigned long long distance(uint64_t a, uint64_t b) {
int biterror = 0;
for (int i = 0; i < 64; i++) {
if (a[i] != b[i]) biterror++;
@@ -427,7 +427,7 @@ public:
class DCTMedian : public DCTDescriptor {
public:
bitset<64> feature(Image* img) {
uint64_t feature(Image* img) {
Image* local_image = new Image(*img);
double* converted = CalculateDCT(local_image);
double bullshit[64];
@@ -437,16 +437,16 @@ public:
sort(bullshit, bullshit + 64);
float median = (float)(bullshit[31] + bullshit[32]) / 2;
bitset<64> retval(0);
uint64_t retval(0);
for (ssize_t i = 0; i < 64; i++) {
retval <<= 1;
if (converted[i] > median)
retval |= bitset<64>(1);
retval |= uint64_t(1);
}
delete local_image;
return retval;
}
unsigned long long distance(bitset<64> a, bitset<64> b) {
unsigned long long distance(uint64_t a, uint64_t b) {
int biterror = 0;
for (int i = 0; i < 64; i++) {
if (a[i] != b[i]) biterror++;
@@ -456,4 +456,4 @@ public:
string ToString() {
return "DCTMedian";
}
};
};*/
DescriptorManager.cpp
+27 -34
View File
@@ -10,13 +10,13 @@ DescriptorManager::DescriptorManager(const fs::path & image_path) {
img_path = image_path;
//Manual definitions.. SAD!
descriptor_types[7] = new DCTMedian();
/*descriptor_types[7] = new DCTMedian();
descriptor_types[6] = new DCTPearson();
descriptor_types[5] = new threshold();
descriptor_types[4] = new pearson();
descriptor_types[3] = new sobel();
descriptor_types[2] = new median();
descriptor_types[1] = new numColors();
descriptor_types[1] = new numColors();*/
descriptor_types[0] = new averageColor();
mutators[0] = new Grayscale();
@@ -47,7 +47,7 @@ void DescriptorManager::Init() {
}
fs::ifstream ifs;
fs::ofstream ofs;
LoadFeatures();
//LoadFeatures();
numImages = std::count_if(
fs::directory_iterator(img_path),
@@ -56,17 +56,18 @@ void DescriptorManager::Init() {
//foreach image in directory (non-recursive)
int filecount = 0;
for (fs::directory_iterator itr(img_path); itr != fs::directory_iterator(); ++itr) {
cout << "\r" << "Progress: " << filecount*100/numImages << "% - (" << filecount << "/" << numImages << ")" << flush;
if (!itr->path().has_extension()) continue; //skip directories and such
filecount++;
string image_name = itr->path().filename().generic_string();
cout << setw(60) << setfill('*') << "\r" << string(60, ' ') << flush;
cout << setfill('*') << setw(60) << right<< "\rProgress: " << filecount*100/numImages << "% - (" << filecount << "/" << numImages << "): " << image_name << flush;
Image* image = new Image;
image->read(itr->path().string());
//foreach descriptor
for (int i = 0; i < numDescriptors; i++) {
//First check if we already have this information from the loaded cache
if (features[i].count(image_name) > 0) {
cout << "We already have feature for " << image_name << endl;
//cout << "We already have feature for " << image_name << endl;
continue;
}
//folder to store descriptors; create if not exists
@@ -76,24 +77,20 @@ void DescriptorManager::Init() {
//cout << "Starting with descriptor " << i << " on location " << descriptorBase.generic_string() << endl;
fs::path featurePath(descriptorBase / itr->path().stem());
bitset<64>* b1;
uint64_t b1;
if (fs::exists(featurePath)) { //feature on disk, load it in
ifs.open(featurePath, fs::fstream::binary);
unsigned long long n;
ifs.read(reinterpret_cast<char*>(&n), sizeof(n));
ifs.read(reinterpret_cast<char*>(&b1), sizeof(b1));
ifs.close();
b1 = new bitset<64>(n);
}
else { //not on disk, calculate it
b1 = new bitset<64>(descriptor_types[i]->feature(image));
unsigned long long n = b1->to_ullong();
b1 = descriptor_types[i]->feature(image);
ofs.open(featurePath, fs::fstream::binary);
ofs.write(reinterpret_cast<const char*>(&n), sizeof(n));
ofs.write(reinterpret_cast<const char*>(&b1), sizeof(b1));
ofs.close();
}
features[i][image_name] = *b1;
SaveFeatures();
delete b1;
features[i][image_name] = b1;
//SaveFeatures();
}
for (int i = 0; i < numMutations; i++) {
//Check if we have somewhere to store the mutation
@@ -114,7 +111,7 @@ void DescriptorManager::Init() {
mutated = mutators[i]->getMutated(image);
mutated->write(mutPath.generic_string());
auto duration = duration_cast<microseconds>(high_resolution_clock::now() - start);
cout << "Calculating mutation " << mutators[i]->ToString() << " took " << duration.count() << "μs." << endl;
//cout << "Calculating mutation " << mutators[i]->ToString() << " took " << duration.count() << "μs." << endl;
}
else {
@@ -123,7 +120,7 @@ void DescriptorManager::Init() {
mutated = new Image;
mutated->read(mutPath.generic_string());
auto duration = duration_cast<microseconds>(high_resolution_clock::now() - start);
cout << "Loading mutation " << mutators[i]->ToString() << " took " << duration.count() << "μs." << endl;
//cout << "Loading mutation " << mutators[i]->ToString() << " took " << duration.count() << "μs." << endl;
}
//now calculate all the features for the mutated images as well
@@ -136,27 +133,23 @@ void DescriptorManager::Init() {
fs::path featurePath(descriptorBase / itr->path().stem());
//cout << featurePath << endl;
string image_name = itr->path().filename().generic_string();
bitset<64>* b1 = NULL;
uint64_t b1;
if (!fs::exists(featurePath)) { //feature not yet calculated
//descriptor_types[i]->distance(b1, b1);
//descriptors[i][image_name] = b1;
b1 = new bitset<64>(descriptor_types[j]->feature(mutated));
unsigned long long n = b1->to_ullong();
b1 = descriptor_types[j]->feature(mutated);
//cout << "Writing " << *b1 << "(" << n << ")" << " to " << featurePath << endl;
ofs.open(featurePath, fs::fstream::binary); //in binary mode; only open file after calculations have been done
ofs.write(reinterpret_cast<const char*>(&n), sizeof(n));
ofs.write(reinterpret_cast<const char*>(&b1), sizeof(b1));
ofs.close();
}
else {
ifs.open(featurePath, fs::fstream::binary);
unsigned long long n;
ifs.read(reinterpret_cast<char*>(&n), sizeof(n));
ifs.read(reinterpret_cast<char*>(&b1), sizeof(b1));
ifs.close();
b1 = new bitset<64>(n);
}
mutatedFeatures[j][i][image_name] = *b1;
if(b1) delete b1;
mutatedFeatures[j][i][image_name] = b1;
}
delete mutated;
}
@@ -211,7 +204,7 @@ void DescriptorManager::LoadFeatures() {
//Sorts all base (unmutated) images in order of distance from the given image
//After this function the distances array will be filled
void DescriptorManager::rankImages(string image_name, int method) {
bitset<64> target;
uint64_t target;
if (features[method].count(image_name) && initialized) { //if we have image
target = features[method][image_name];
}
@@ -219,7 +212,7 @@ void DescriptorManager::rankImages(string image_name, int method) {
return;
}
distances.clear(); //start fresh
for (map<string, bitset<64>>::iterator it = features[method].begin(); it != features[method].end(); ++it) {
for (map<string, uint64_t>::iterator it = features[method].begin(); it != features[method].end(); ++it) {
unsigned long long d = descriptor_types[method]->distance(target, it->second);
pair<unsigned long long, string> p(d, it->first);
distances.insert(p); //pairs are compared by their first element..
@@ -230,7 +223,7 @@ void DescriptorManager::rankImages(string image_name, int method) {
//It inserts these mutated descriptors in a sorted way
//Assumes the distances array is filled with the normal distances
multiset<pair<unsigned long long, string>> DescriptorManager::rankMutations(string image_name, int method) {
bitset<64> target;
uint64_t target;
if (features[method].count(image_name) && initialized) { //this should always be the case
target = features[method][image_name];
}
@@ -238,7 +231,7 @@ multiset<pair<unsigned long long, string>> DescriptorManager::rankMutations(stri
//cout << "Finding hits for " << image_name << " which has feature-number " << target.to_ullong() << endl;
multiset<pair<unsigned long long, string>> local_distance = distances; //get our local copy of the distances.
for (int i = 0; i < numMutations; i++) {
bitset<64> b1 = mutatedFeatures[method][i][image_name];
uint64_t b1 = mutatedFeatures[method][i][image_name];
unsigned long long d = descriptor_types[method]->distance(target, b1);
pair<unsigned long long, string> p(d, mutators[i]->ToString());
local_distance.insert(p);
@@ -265,7 +258,7 @@ multiset<pair<unsigned long long, string>> DescriptorManager::rankMutations(stri
void DescriptorManager::runExperiments() {
fs::ofstream ofs;
for (int i = 0; i < numDescriptors; i++) {
for (map<string, bitset<64>>::iterator it = features[i].begin(); it != features[i].end(); ++it) {
for (map<string, uint64_t>::iterator it = features[i].begin(); it != features[i].end(); ++it) {
//cout << "=== " << descriptor_types[i]->ToString() << ": " << it->first << " ===" << endl;
rankImages(it->first, i);
multiset<pair<unsigned long long, string>> local_distance = rankMutations(it->first, i);
@@ -300,13 +293,13 @@ void DescriptorManager::runExperiments() {
//assumes a < numDescriptors
void DescriptorManager::outputMap(int a) {
cout << "Outputting map " << descriptor_types[a]->ToString() << endl;
map<string, bitset<64>> a_map = features[a];
map<string, uint64_t> a_map = features[a];
if (a_map.begin() == a_map.end()) {
cout << "------------------------" << endl << "No data found" << endl << endl;
}
for (map<string, bitset<64>>::iterator it = a_map.begin(); it != a_map.end(); ++it) {
cout << it->first << " => " << it->second << " (" << it->second.to_ullong() << ")" << endl;
for (map<string, uint64_t>::iterator it = a_map.begin(); it != a_map.end(); ++it) {
cout << it->first << " => " << it->second << " (" << bitset<64>(it->second) << ")" << endl;
}
}
DescriptorManager.h
+3 -3
View File
@@ -54,15 +54,15 @@ public:
void outputMap();
void outputRank(string imagename, int method);
private:
static const int numDescriptors = 8;
static const int numDescriptors = 1;
static const int numMutations = 8;
//TODO: automagically update this?
bool initialized = false;
int numImages;
map<string, bitset<64>> features[numDescriptors]; //array of maps that will hold the descriptors
map<string, bitset<64>> mutatedFeatures[numDescriptors][numMutations];
map<string, uint64_t> features[numDescriptors]; //array of maps that will hold the descriptors
map<string, uint64_t> mutatedFeatures[numDescriptors][numMutations];
multiset<pair<unsigned long long, string> > distances; //multiset that stores the distances
fs::path img_path; //where the images are
Descriptor* descriptor_types[numDescriptors];
main.cpp
+3 -1
View File
@@ -43,8 +43,10 @@ int main(int argc, char **argv) {
//smanager->generatePreviews(imagename);
//cout << "Intializing..." << endl;
manager->Init();
cout << endl;
manager->outputMap();
//cout << "done!" << endl;
manager->runExperiments();
// manager->runExperiments();
return 0;
}