00001
00020
00021
00022
00023 #include "metrics.h"
00024 #include "bestfirst.h"
00025 #include "associate.h"
00026 #include "tally.h"
00027 #include "plotseg.h"
00028 #include "globals.h"
00029 #include "wordclass.h"
00030 #include "intmatcher.h"
00031 #include "freelist.h"
00032 #include "djmenus.h"
00033 #include "callcpp.h"
00034
00035
00036
00037
00039 static int states_timed_out1;
00040 static int states_timed_out2;
00041 static int words_segmented1;
00042 static int words_segmented2;
00043 static int segmentation_states1;
00044 static int segmentation_states2;
00045 static int save_priorities;
00046
00047 int words_chopped1;
00048 int words_chopped2;
00049 int chops_attempted1;
00050 int chops_performed1;
00051 int chops_attempted2;
00052 int chops_performed2;
00053
00054 int character_count;
00055 int word_count;
00057 int chars_classified;
00058
00059 MEASUREMENT num_pieces;
00060 MEASUREMENT width_measure;
00061
00063 MEASUREMENT width_priority_range;
00064 MEASUREMENT match_priority_range;
00065
00066 TALLY states_before_best;
00067 TALLY best_certainties[2];
00069 TALLY character_widths;
00070
00072 FILE *priority_file_1;
00073 FILE *priority_file_2;
00074 FILE *priority_file_3;
00075
00077 STATE *known_best_state = NULL;
00078
00079
00080
00081
00082 #define CERTAINTY_BUCKET_SIZE -0.5
00083 #define CERTAINTY_BUCKETS 40
00084
00085
00086
00087
00088
00095 void init_metrics() {
00096 words_chopped1 = 0;
00097 words_chopped2 = 0;
00098 chops_performed1 = 0;
00099 chops_performed2 = 0;
00100 chops_attempted1 = 0;
00101 chops_attempted2 = 0;
00102
00103 words_segmented1 = 0;
00104 words_segmented2 = 0;
00105 states_timed_out1 = 0;
00106 states_timed_out2 = 0;
00107 segmentation_states1 = 0;
00108 segmentation_states2 = 0;
00109
00110 save_priorities = 0;
00111
00112 character_count = 0;
00113 word_count = 0;
00114 chars_classified = 0;
00115 permutation_count = 0;
00116
00117 states_before_best = new_tally (min (100, num_seg_states));
00118
00119 best_certainties[0] = new_tally (CERTAINTY_BUCKETS);
00120 best_certainties[1] = new_tally (CERTAINTY_BUCKETS);
00121 }
00122
00123
00128 void end_metrics() {
00129 memfree(states_before_best);
00130 memfree(best_certainties[0]);
00131 memfree(best_certainties[1]);
00132 memfree(character_widths);
00133 states_before_best = NULL;
00134 best_certainties[0] = NULL;
00135 best_certainties[1] = NULL;
00136 character_widths = NULL;
00137 }
00138
00139
00140
00145 void record_certainty(float certainty, int pass) {
00146 int bucket;
00147
00148 if (certainty / CERTAINTY_BUCKET_SIZE < MAXINT)
00149 bucket = (int) (certainty / CERTAINTY_BUCKET_SIZE);
00150 else
00151 bucket = MAXINT;
00152
00153 inc_tally_bucket (best_certainties[pass - 1], bucket);
00154 }
00155
00156
00157
00164 void record_search_status(int num_states, int before_best, float closeness) {
00165 inc_tally_bucket(states_before_best, before_best);
00166
00167 if (first_pass) {
00168 if (num_states == num_seg_states + 1)
00169 states_timed_out1++;
00170 segmentation_states1 += num_states;
00171 words_segmented1++;
00172 }
00173 else {
00174 if (num_states == num_seg_states + 1)
00175 states_timed_out2++;
00176 segmentation_states2 += num_states;
00177 words_segmented2++;
00178 }
00179 }
00180
00181
00182
00186 void save_summary(INT32 elapsed_time) {
00187 #ifndef SECURE_NAMES
00188 char outfilename[CHARS_PER_LINE];
00189 FILE *f;
00190 int x;
00191 int total;
00192
00193 strcpy(outfilename, imagefile);
00194 strcat (outfilename, ".sta");
00195 f = open_file (outfilename, "w");
00196
00197 fprintf (f, INT32FORMAT " seconds elapsed\n", elapsed_time);
00198 fprintf (f, "\n");
00199
00200 fprintf (f, "%d characters\n", character_count);
00201 fprintf (f, "%d words\n", word_count);
00202 fprintf (f, "\n");
00203
00204 fprintf (f, "%d permutations performed\n", permutation_count);
00205 fprintf (f, "%d characters classified\n", chars_classified);
00206 fprintf (f, "%4.0f%% classification overhead\n",
00207 (float) chars_classified / character_count * 100.0 - 100.0);
00208 fprintf (f, "\n");
00209
00210 fprintf (f, "%d words chopped (pass 1) ", words_chopped1);
00211 fprintf (f, " (%0.0f%%)\n", (float) words_chopped1 / word_count * 100);
00212 fprintf (f, "%d chops performed\n", chops_performed1);
00213 fprintf (f, "%d chops attempted\n", chops_attempted1);
00214 fprintf (f, "\n");
00215
00216 fprintf (f, "%d words joined (pass 1)", words_segmented1);
00217 fprintf (f, " (%0.0f%%)\n", (float) words_segmented1 / word_count * 100);
00218 fprintf (f, "%d segmentation states\n", segmentation_states1);
00219 fprintf (f, "%d segmentations timed out\n", states_timed_out1);
00220 fprintf (f, "\n");
00221
00222 fprintf (f, "%d words chopped (pass 2) ", words_chopped2);
00223 fprintf (f, " (%0.0f%%)\n", (float) words_chopped2 / word_count * 100);
00224 fprintf (f, "%d chops performed\n", chops_performed2);
00225 fprintf (f, "%d chops attempted\n", chops_attempted2);
00226 fprintf (f, "\n");
00227
00228 fprintf (f, "%d words joined (pass 2)", words_segmented2);
00229 fprintf (f, " (%0.0f%%)\n", (float) words_segmented2 / word_count * 100);
00230 fprintf (f, "%d segmentation states\n", segmentation_states2);
00231 fprintf (f, "%d segmentations timed out\n", states_timed_out2);
00232 fprintf (f, "\n");
00233
00234 total = 0;
00235 iterate_tally (states_before_best, x)
00236 total += (tally_entry (states_before_best, x) * x);
00237 fprintf (f, "segmentations (before best) = %d\n", total);
00238 if (total != 0.0)
00239 fprintf (f, "%4.0f%% segmentation overhead\n",
00240 (float) (segmentation_states1 + segmentation_states2) /
00241 total * 100.0 - 100.0);
00242 fprintf (f, "\n");
00243
00244 print_tally (f, "segmentations (before best)", states_before_best);
00245
00246 iterate_tally (best_certainties[0], x)
00247 cprintf ("best certainty of %8.4f = %4d %4d\n",
00248 x * CERTAINTY_BUCKET_SIZE,
00249 tally_entry (best_certainties[0], x),
00250 tally_entry (best_certainties[1], x));
00251
00252 PrintIntMatcherStats(f);
00253 dj_statistics(f);
00254 fclose(f);
00255 #endif
00256 }
00257
00258
00259
00267 void record_priorities(SEARCH_RECORD *the_search,
00268 STATE *old_state,
00269 FLOAT32 priority_1,
00270 FLOAT32 priority_2) {
00271 record_samples(priority_1, priority_2);
00272 }
00273
00274
00275
00279 void record_samples(FLOAT32 match_pri, FLOAT32 width_pri) {
00280 ADD_SAMPLE(match_priority_range, match_pri);
00281 ADD_SAMPLE(width_priority_range, width_pri);
00282 }
00283
00284
00285
00289 void reset_width_tally() {
00290 character_widths = new_tally (20);
00291 new_measurement(width_measure);
00292 width_measure.num_samples = 158;
00293 width_measure.sum_of_samples = 125.0;
00294 width_measure.sum_of_squares = 118.0;
00295 }
00296
00297
00298
00302 #ifndef GRAPHICS_DISABLED
00303 void save_best_state(CHUNKS_RECORD *chunks_record) {
00304 STATE state;
00305 SEARCH_STATE chunk_groups;
00306 int num_joints;
00307
00308 if (save_priorities) {
00309 num_joints = matrix_dimension (chunks_record->ratings) - 1;
00310
00311 state.part1 = 0xffffffff;
00312 state.part2 = 0xffffffff;
00313
00314 chunk_groups = bin_to_chunks (&state, num_joints);
00315 display_segmentation (chunks_record->chunks, chunk_groups);
00316 memfree(chunk_groups);
00317
00318 cprintf ("Enter the correct segmentation > ");
00319 fflush(stdout);
00320 state.part1 = 0;
00321 scanf ("%x", &state.part2);
00322
00323 chunk_groups = bin_to_chunks (&state, num_joints);
00324 display_segmentation (chunks_record->chunks, chunk_groups);
00325 memfree(chunk_groups);
00326 window_wait(segm_window);
00327
00328 if (known_best_state)
00329 free_state(known_best_state);
00330 known_best_state = new_state (&state);
00331 }
00332 }
00333 #endif
00334
00335
00336
00340 void start_recording() {
00341 if (save_priorities) {
00342 priority_file_1 = open_file ("Priorities1", "w");
00343 priority_file_2 = open_file ("Priorities2", "w");
00344 priority_file_3 = open_file ("Priorities3", "w");
00345 }
00346 }
00347
00348
00349
00353 void stop_recording() {
00354 if (save_priorities) {
00355 fclose(priority_file_1);
00356 fclose(priority_file_2);
00357 fclose(priority_file_3);
00358 }
00359 }