Changeset 97

Timestamp:

Apr 6, 2010, 7:16:33 AM (15 years ago)

Author:

Rick van der Zwet

Message:

New classifier block detection. Not yet optimized for blocks

Location:

liacs/MIR2010/SourceCode

Files:

• hvdzwet.vcproj (modified) (1 diff)
• main.cpp (modified) (21 diffs)

liacs/MIR2010/SourceCode/hvdzwet.vcproj

@@ -256 +256 @@
                         </File>
                         <File
+                                RelativePath=".\reference_implementation.h"
+                                >
+                        </File>
+                        <File
                                 RelativePath="XBrowseForFolder.h"
                                 >

liacs/MIR2010/SourceCode/main.cpp

@@ -40 +40 @@
 float average[CATEGORY_SIZE][BIN_COUNT*BIN_COUNT*BIN_COUNT];
 
+
+// Find common coloured shapes as characteristics
+// Walk over image with SPREAD steps, if pixel -within TOLERANCE- matches previous pixel 
+// make the size of the specific block bigger.
+#define TOLERANCE 20
+#define SPREAD 20
+#define MAX_SIZE 500
+float average_block[CATEGORY_SIZE][MAX_SIZE/SPREAD];
+
+
 // Some prototyping foo on functions
+#define p_err(err_msg) printf("ERROR: %s\n", err_msg);
 #define _return(err_msg, retval) printf("DEBUG: %s\n",err_msg); cin.get(); return(retval);
-
-bool verbose = false;
+#define cout_status(msg,flag) cout << msg; (flag) ? cout << " off" : cout << " on"; cout << endl;
+
+
+bool opt_verbose = false;
+bool opt_histogram = true;
+bool opt_block = true;
 
 bool file_exists(const char * filename)
@@ -114 +129 @@
                 cout << "3. determine winter sport" << endl;
                 cout << "4. batch test winter sport" << endl;
-                cout << "v. Turn verbose mode ";
-                if (verbose)
-                        cout << "off" << endl;
-                else
-                        cout << "on" << endl;
+                cout_status("v. Turn verbose mode", opt_verbose);
+                cout_status("h. Turn histogram classifier", opt_histogram);
+                cout_status("b. Turn block classifier", opt_block);
+
                 cout << endl;
                 cout << "Please select option, or type 'q' to quit: ";
@@ -126 +140 @@
                 // start the chosen option
                 switch (c)
-                {
+                { 
                 case 'q':
                         return 0;
@@ -158 +172 @@
                         break;
                 case 'v':
-                        verbose = (!verbose);
+                        opt_verbose = (!opt_verbose);
+                        break;
+                case 'b':
+                        opt_block = (!opt_block);
+                        break;
+                case 'h':
+                        opt_histogram = (!opt_histogram);
+                        break;
                 default:
                         continue;
@@ -207 +228 @@
 }
 
+// histogram should be a preallocated array of size BIN_COUNT*BIN_COUNT*BIN_COUNT elements and will
+// be filled with the color histogram of the image where path points at
+bool CalculateBlock(const char *path, float *block)
+{
+        // load the image
+        CxImage image(path, CXIMAGE_FORMAT_JPG);
+        if (!image.IsValid())
+                return false;
+        // clear histogram
+        memset(block, 0, MAX_SIZE/SPREAD * sizeof(float));
+        // walk through the pixels to fill the histogram
+        const int width = (int)image.GetWidth();
+        const int height = (int)image.GetHeight();
+
+        int rgb_value = 0;
+        int rgb_prev = 0;
+
+        int block_size = 0;
+
+        for (int y = 0; y < height; y += SPREAD)
+        {
+                for (int x = 0; x < width; x += SPREAD)
+                {
+                        // Note: CxImage library starts counting at lower-left corner of the image,
+                        //       which is seen as the top of the image.
+                        RGBQUAD rgb = image.BlindGetPixelColor(x, y, false);
+                        rgb_value = (rgb.rgbRed + rgb.rgbBlue + rgb.rgbGreen);
+
+                        if (abs(rgb_value - rgb_prev) > TOLERANCE) {
+                                block[block_size]++;
+                                block_size = 1;
+                        }
+                        rgb_prev = rgb_value;
+                }
+        }
+        return true;
+}
+
 bool CalculateDescriptors(const char *basedir)
 {
         // the histogram that we reuse for each image
         float *histogram = new float[BIN_COUNT*BIN_COUNT*BIN_COUNT];
+        float *block = new float[MAX_SIZE/SPREAD];
         // walk through all images
         // Note: each of the three categories has 50 images
@@ -234 +294 @@
                         if (!CalculateDescriptor(path, histogram))
                                 goto failure;
-                        // save the descriptor to disk
+
+                        if (!CalculateBlock(path, block))
+                                goto failure;
+
+                        // save the descriptor,block to disk
                         SAFE_SPRINTF(path, sizeof(path), "%s%i.dat", catdir, i);
                         if ((file = fopen(path, "wb")) == NULL)
@@ -240 +304 @@
                         if (fwrite(histogram, sizeof(float), BIN_COUNT*BIN_COUNT*BIN_COUNT, file) != BIN_COUNT*BIN_COUNT*BIN_COUNT)
                                 goto failure;
+                        if (fwrite(block, sizeof(float), MAX_SIZE/SPREAD, file) != MAX_SIZE/SPREAD)
+                                goto failure;
                         SAFE_CLOSEFILE(file);
+
+
                 }
         }
         // release resources
         SAFE_DELETE_ARRAY(histogram);
+        SAFE_DELETE_ARRAY(block);
         return true;
 
@@ -250 +319 @@
         SAFE_CLOSEFILE(file);
         SAFE_DELETE_ARRAY(histogram);
+        SAFE_DELETE_ARRAY(block);
         return false;
 }
@@ -258 +328 @@
         // characteristics of that category
         float *histogram = new float[BIN_COUNT*BIN_COUNT*BIN_COUNT];
+        float *block = new float[MAX_SIZE/SPREAD];
+        float *average_block = new float[MAX_SIZE/SPREAD];
         float *average = new float[BIN_COUNT*BIN_COUNT*BIN_COUNT];
         // walk through all descriptors
@@ -277 +349 @@
                         SAFE_SPRINTF(path, sizeof(path), "%s%i.dat", catdir, i);
                         if (!file_exists(path)) {
+                                p_err("File does not exists");
                                 continue;
                         }
                         cout << "[" << catname << "] processing image " << i << endl;
                         // load the histogram descriptor
-                        if ((file = fopen(path, "rb")) == NULL)
-                                goto failure;
-                        if (fread(histogram, sizeof(float), BIN_COUNT*BIN_COUNT*BIN_COUNT, file) != BIN_COUNT*BIN_COUNT*BIN_COUNT)
-                                goto failure;
+                        if ((file = fopen(path, "rb")) == NULL) {
+                                p_err("Cannot open average datafile");
+                                goto failure;
+                        }
+                        if (fread(histogram, sizeof(float), BIN_COUNT*BIN_COUNT*BIN_COUNT, file) != BIN_COUNT*BIN_COUNT*BIN_COUNT) {
+                                p_err("Cannot read histogram");
+                                goto failure;
+                        }
+                        if (fread(block, sizeof(float), MAX_SIZE/SPREAD, file) != MAX_SIZE/SPREAD) {
+                                p_err("Cannot read block");
+                                goto failure;
+                        }
+
                         SAFE_CLOSEFILE(file);
                         // add the value of each bin to the average
@@ -290 +372 @@
                                 average[b] += histogram[b];
 
+                        for (int b = 0; b < MAX_SIZE/SPREAD; b++)
+                                average_block[b] += block[b];
+
                         c_size++;
                 }
@@ -295 +380 @@
                 for (int b = 0; b < BIN_COUNT*BIN_COUNT*BIN_COUNT; b++)
                         average[b] /= c_size;
+                
+                for (int b = 0; b < MAX_SIZE/SPREAD; b++)
+                        average_block[b] /= c_size;
+
                 // save the average to disk
                 SAFE_SPRINTF(path, sizeof(path), "%s%s.dat", catdir, "average");
@@ -301 +390 @@
                 if (fwrite(average, sizeof(float), BIN_COUNT*BIN_COUNT*BIN_COUNT, file) != BIN_COUNT*BIN_COUNT*BIN_COUNT)
                         goto failure;
+                if (fwrite(average_block, sizeof(float), MAX_SIZE/SPREAD, file) != MAX_SIZE/SPREAD)
+                        goto failure;
+
                 SAFE_CLOSEFILE(file);
         }
@@ -306 +398 @@
         SAFE_DELETE_ARRAY(histogram);
         SAFE_DELETE_ARRAY(average);
+        SAFE_DELETE_ARRAY(block);
         return true;
 
@@ -311 +404 @@
         SAFE_CLOSEFILE(file);
         SAFE_DELETE_ARRAY(histogram);
-        SAFE_DELETE_ARRAY(average);
+        SAFE_DELETE_ARRAY(block);
         return false;
 }
@@ -335 +428 @@
                 if (fread(average[c], sizeof(float), BIN_COUNT*BIN_COUNT*BIN_COUNT, file) != BIN_COUNT*BIN_COUNT*BIN_COUNT)
                         return false;
+                if (fread(average_block[c], sizeof(float), MAX_SIZE/SPREAD, file) != MAX_SIZE/SPREAD)
+                        return false;
+
                 SAFE_CLOSEFILE(file);
         }
@@ -340 +436 @@
 }
 
-int DetermineCategory(const char *path, const int guess=-1) {
+int DetermineCategory(const char *path, const int guess=-1, const bool verbose=false) {
         float *histogram = new float[BIN_COUNT*BIN_COUNT*BIN_COUNT];
+        float *block = new float[MAX_SIZE/SPREAD];
         float cat2dist[CATEGORY_SIZE];
+        float cat2block[CATEGORY_SIZE];
 
         /* First category default best canidate */
-        int retval = 0;
+        int cat_histogram = 0;
+        int cat_block = 0;
 
         /* calculate the histogram of the image */
@@ -351 +450 @@
                 return -1;
 
+        if (!CalculateBlock(path, block))
+                return -1;
+
         /* determine the distance to each category */
         for (int c = 0; c < CATEGORY_SIZE; c++)
@@ -356 +458 @@
                 // determine distance
                 cat2dist[c] = 0.0f;
+                cat2block[c] = 0.0f;
                 for (int b = 0; b < BIN_COUNT*BIN_COUNT*BIN_COUNT; b++)
                         cat2dist[c] += fabs(histogram[b] - average[c][b]);
+
+                for (int b = 0; b < MAX_SIZE/SPREAD; b++)
+                        cat2block[c] += fabs(block[b] - average_block[c][b]);
         }
 
         /* determine the winning category */
-        for (int i = 1; i < CATEGORY_SIZE; i++)
-                if (cat2dist[i] < cat2dist[retval])
-                        retval = i;
+        for (int i = 1; i < CATEGORY_SIZE; i++) {
+                if (cat2dist[i] < cat2dist[cat_histogram])
+                        cat_histogram = i;
+                if (cat2block[i] < cat2block[cat_block])
+                        cat_block = i;
+        }
 
         if (verbose) {
                 /* Dirty hack to show some more details in case of failure */
-                if (guess != -1 && guess != retval) {
+                if (opt_histogram && guess != -1 && guess != cat_histogram) {
                         for (int i = 0; i < CATEGORY_SIZE; i++) {               
-                                printf("%s Distance to %-20s: %f %s\n", (retval == i) ? "*" : " ", 
-                                        categories[i], cat2dist[i],(retval == i) ? "*" : "");
-                        }
-                }
+                                printf("%s [histogram] distance to %-20s: %f %s\n", (cat_histogram == i) ? "*" : " ", 
+                                        categories[i], cat2dist[i],(cat_histogram == i) ? "*" : "");
+                        }
+                }
+                if (opt_block && guess != -1 && guess != cat_block) {
+                        for (int i = 0; i < CATEGORY_SIZE; i++) {               
+                                printf("%s [block] distance to %-20s: %f %s\n", (cat_block == i) ? "*" : " ", 
+                                        categories[i], cat2block[i],(cat_block == i) ? "*" : "");
+                        }
+                }
+
         }
 
         /* return result */
-        return retval;
+        if (opt_histogram) {
+                return cat_histogram;
+        }
+        else if (opt_block) {
+                return cat_block;
+        } else {
+            return -1;
+        }
 }
 
@@ -452 +575 @@
                                 c_succes[c]++;
                         } else {
-                                cout << "[" << catname << "] testing image " << i << " : FAIL" << endl;                 }
+                                cout << "[" << catname << "] testing image " << i << " : FAIL" << endl;
+                                DetermineCategory(path,c,opt_verbose);
+                        }
                 }
                 cout << "[" << catname << "] results " << c_succes[c] << "/" << c_total[c] << endl;
@@ -459 +584 @@
         /* Display grand total */
         cout << "=== Totals ===" << endl;
+        cout << "Clasifier used: ";
+        if (opt_histogram)
+                cout << "histogram";
+        else if (opt_block)
+                cout << "block";
+        cout << endl;
+
         for (int c = 0; c < CATEGORY_SIZE; c++)
         {