/// Builds a PAGE_RES from the block_list in the way required for ApplyBoxes:
/// All fuzzy spaces are removed, and all the words are maximally chopped.
PAGE_RES* Tesseract::SetupApplyBoxes(const GenericVector<TBOX>& boxes,
BLOCK_LIST *block_list) {
PreenXHeights(block_list);
// Strip all fuzzy space markers to simplify the PAGE_RES.
BLOCK_IT b_it(block_list);
for (b_it.mark_cycle_pt(); !b_it.cycled_list(); b_it.forward()) {
BLOCK* block = b_it.data();
ROW_IT r_it(block->row_list());
for (r_it.mark_cycle_pt(); !r_it.cycled_list(); r_it.forward ()) {
ROW* row = r_it.data();
WERD_IT w_it(row->word_list());
for (w_it.mark_cycle_pt(); !w_it.cycled_list(); w_it.forward()) {
WERD* word = w_it.data();
if (word->cblob_list()->empty()) {
delete w_it.extract();
} else {
word->set_flag(W_FUZZY_SP, false);
word->set_flag(W_FUZZY_NON, false);
}
}
}
}
PAGE_RES* page_res = new PAGE_RES(false, block_list, NULL);
PAGE_RES_IT pr_it(page_res);
WERD_RES* word_res;
while ((word_res = pr_it.word()) != NULL) {
MaximallyChopWord(boxes, pr_it.block()->block,
pr_it.row()->row, word_res);
pr_it.forward();
}
return page_res;
}
PAGE_RES_IT* make_pseudo_word(PAGE_RES* page_res, const TBOX& selection_box) {
PAGE_RES_IT pr_it(page_res);
C_BLOB_LIST new_blobs; // list of gathered blobs
C_BLOB_IT new_blob_it = &new_blobs; // iterator
for (WERD_RES* word_res = pr_it.word(); word_res != NULL;
word_res = pr_it.forward()) {
WERD* word = word_res->word;
if (word->bounding_box().overlap(selection_box)) {
C_BLOB_IT blob_it(word->cblob_list());
for (blob_it.mark_cycle_pt();
!blob_it.cycled_list(); blob_it.forward()) {
C_BLOB* blob = blob_it.data();
if (blob->bounding_box().overlap(selection_box)) {
new_blob_it.add_after_then_move(C_BLOB::deep_copy(blob));
}
}
if (!new_blobs.empty()) {
WERD* pseudo_word = new WERD(&new_blobs, 1, NULL);
word_res = pr_it.InsertSimpleCloneWord(*word_res, pseudo_word);
PAGE_RES_IT* it = new PAGE_RES_IT(page_res);
while (it->word() != word_res && it->word() != NULL) it->forward();
ASSERT_HOST(it->word() == word_res);
return it;
}
}
}
return NULL;
}
void
SOP_PrimGroupCentroid::baryCenter(const GU_Detail *input_geo,
GA_Range &pr_range,
const GA_PrimitiveList &prim_list,
UT_Vector3 &pos)
{
GA_Range pt_range;
GA_OffsetArray points;
GA_OffsetArray::const_iterator points_it;
// We need to iterate over each primitive in the range and
// find out which points it references.
for (GA_Iterator pr_it(pr_range); !pr_it.atEnd(); ++pr_it)
{
// Get the range of points for the primitive using the
// offset from the primitive list.
pt_range = prim_list.get(*pr_it)->getPointRange();
// Add each point's offset to the array, checking for duplicates.
for (GA_Iterator pt_it(pt_range); !pt_it.atEnd(); ++pt_it)
points.append(*pt_it, true);
}
// Reset the position.
pos.assign(0,0,0);
// Add the positions for all the points.
for (points_it = points.begin(); !points_it.atEnd(); ++points_it)
pos += input_geo->getPos3(*points_it);
// Store the average position for all the points we found.
pos /= points.entries();
}
void
SOP_PrimGroupCentroid::boundingBox(const GU_Detail *input_geo,
GA_Range &pr_range,
const GA_PrimitiveList &prim_list,
UT_Vector3 &pos)
{
GA_Range pt_range;
UT_BoundingBox bbox;
// Initialize the bounding box to contain nothing and have
// no position.
bbox.initBounds();
// Iterate over each primitive in the range.
for (GA_Iterator pr_it(pr_range); !pr_it.atEnd(); ++pr_it)
{
// Get the range of points for the primitive using the
// offset from the primitive list.
pt_range = prim_list.get(*pr_it)->getPointRange();
// For each point in the primitive, enlarge the bounding
// box to contain it.
for (GA_Iterator pt_it(pt_range); !pt_it.atEnd(); ++pt_it)
bbox.enlargeBounds(input_geo->getPos3(*pt_it));
}
// Extract the center.
pos = bbox.center();
}
/// - Counts up the labelled words and the blobs within.
/// - Deletes all unused or emptied words, counting the unused ones.
/// - Resets W_BOL and W_EOL flags correctly.
/// - Builds the rebuild_word and rebuilds the box_word and the best_choice.
void Tesseract::TidyUp(PAGE_RES* page_res) {
int ok_blob_count = 0;
int bad_blob_count = 0;
int ok_word_count = 0;
int unlabelled_words = 0;
PAGE_RES_IT pr_it(page_res);
WERD_RES* word_res;
for (; (word_res = pr_it.word()) != NULL; pr_it.forward()) {
int ok_in_word = 0;
int blob_count = word_res->correct_text.size();
WERD_CHOICE* word_choice = new WERD_CHOICE(word_res->uch_set, blob_count);
word_choice->set_permuter(TOP_CHOICE_PERM);
for (int c = 0; c < blob_count; ++c) {
if (word_res->correct_text[c].length() > 0) {
++ok_in_word;
}
// Since we only need a fake word_res->best_choice, the actual
// unichar_ids do not matter. Which is fortunate, since TidyUp()
// can be called while training Tesseract, at the stage where
// unicharset is not meaningful yet.
word_choice->append_unichar_id_space_allocated(
INVALID_UNICHAR_ID, word_res->best_state[c], 1.0f, -1.0f);
}
if (ok_in_word > 0) {
ok_blob_count += ok_in_word;
bad_blob_count += word_res->correct_text.size() - ok_in_word;
word_res->LogNewRawChoice(word_choice);
word_res->LogNewCookedChoice(1, false, word_choice);
} else {
++unlabelled_words;
if (applybox_debug > 0) {
tprintf("APPLY_BOXES: Unlabelled word at :");
word_res->word->bounding_box().print();
}
pr_it.DeleteCurrentWord();
delete word_choice;
}
}
pr_it.restart_page();
for (; (word_res = pr_it.word()) != NULL; pr_it.forward()) {
// Denormalize back to a BoxWord.
word_res->RebuildBestState();
word_res->SetupBoxWord();
word_res->word->set_flag(W_BOL, pr_it.prev_row() != pr_it.row());
word_res->word->set_flag(W_EOL, pr_it.next_row() != pr_it.row());
}
if (applybox_debug > 0) {
tprintf(" Found %d good blobs.\n", ok_blob_count);
if (bad_blob_count > 0) {
tprintf(" Leaving %d unlabelled blobs in %d words.\n",
bad_blob_count, ok_word_count);
}
if (unlabelled_words > 0)
tprintf(" %d remaining unlabelled words deleted.\n", unlabelled_words);
}
}
/// Calls #LearnWord to extract features for labelled blobs within each word.
/// Features are stored in an internal buffer.
void Tesseract::ApplyBoxTraining(const STRING& fontname, PAGE_RES* page_res) {
PAGE_RES_IT pr_it(page_res);
int word_count = 0;
for (WERD_RES *word_res = pr_it.word(); word_res != NULL;
word_res = pr_it.forward()) {
LearnWord(fontname.string(), word_res);
++word_count;
}
tprintf("Generated training data for %d words\n", word_count);
}
/**
* do_re_display()
*
* Redisplay page
*/
void Tesseract::do_re_display(
BOOL8 (tesseract::Tesseract::*word_painter)(PAGE_RES_IT* pr_it)) {
int block_count = 1;
image_win->Clear();
if (display_image != 0) {
image_win->Image(pix_binary_, 0, 0);
}
PAGE_RES_IT pr_it(current_page_res);
for (WERD_RES* word = pr_it.word(); word != NULL; word = pr_it.forward()) {
(this->*word_painter)(&pr_it);
if (display_baselines && pr_it.row() != pr_it.prev_row())
pr_it.row()->row->plot_baseline(image_win, ScrollView::GREEN);
if (display_blocks && pr_it.block() != pr_it.prev_block())
pr_it.block()->block->plot(image_win, block_count++, ScrollView::RED);
}
image_win->Update();
}
/** Creates a fake best_choice entry in each WERD_RES with the correct text.*/
void Tesseract::CorrectClassifyWords(PAGE_RES* page_res) {
PAGE_RES_IT pr_it(page_res);
for (WERD_RES *word_res = pr_it.word(); word_res != NULL;
word_res = pr_it.forward()) {
WERD_CHOICE* choice = new WERD_CHOICE(word_res->uch_set,
word_res->correct_text.size());
for (int i = 0; i < word_res->correct_text.size(); ++i) {
// The part before the first space is the real ground truth, and the
// rest is the bounding box location and page number.
GenericVector<STRING> tokens;
word_res->correct_text[i].split(' ', &tokens);
UNICHAR_ID char_id = unicharset.unichar_to_id(tokens[0].string());
choice->append_unichar_id_space_allocated(char_id,
word_res->best_state[i],
0.0f, 0.0f);
}
word_res->ClearWordChoices();
word_res->LogNewRawChoice(choice);
word_res->LogNewCookedChoice(1, false, choice);
}
}
void show_point(PAGE_RES* page_res, float x, float y) {
FCOORD pt(x, y);
PAGE_RES_IT pr_it(page_res);
const int kBufsize = 512;
char msg[kBufsize];
char *msg_ptr = msg;
msg_ptr += sprintf(msg_ptr, "Pt:(%0.3f, %0.3f) ", x, y);
for (WERD_RES* word = pr_it.word(); word != NULL; word = pr_it.forward()) {
if (pr_it.row() != pr_it.prev_row() &&
pr_it.row()->row->bounding_box().contains(pt)) {
msg_ptr += sprintf(msg_ptr, "BL(x)=%0.3f ",
pr_it.row()->row->base_line(x));
}
if (word->word->bounding_box().contains(pt)) {
TBOX box = word->word->bounding_box();
msg_ptr += sprintf(msg_ptr, "Wd(%d, %d)/(%d, %d) ",
box.left(), box.bottom(),
box.right(), box.top());
C_BLOB_IT cblob_it(word->word->cblob_list());
for (cblob_it.mark_cycle_pt();
!cblob_it.cycled_list();
cblob_it.forward()) {
C_BLOB* cblob = cblob_it.data();
box = cblob->bounding_box();
if (box.contains(pt)) {
msg_ptr += sprintf(msg_ptr,
"CBlb(%d, %d)/(%d, %d) ",
box.left(), box.bottom(),
box.right(), box.top());
}
}
}
}
image_win->AddMessage(msg);
}
/// Resegments the words by running the classifier in an attempt to find the
/// correct segmentation that produces the required string.
void Tesseract::ReSegmentByClassification(PAGE_RES* page_res) {
PAGE_RES_IT pr_it(page_res);
WERD_RES* word_res;
for (; (word_res = pr_it.word()) != NULL; pr_it.forward()) {
WERD* word = word_res->word;
if (word->text() == NULL || word->text()[0] == '\0')
continue; // Ignore words that have no text.
// Convert the correct text to a vector of UNICHAR_ID
GenericVector<UNICHAR_ID> target_text;
if (!ConvertStringToUnichars(word->text(), &target_text)) {
tprintf("APPLY_BOX: FAILURE: can't find class_id for '%s'\n",
word->text());
pr_it.DeleteCurrentWord();
continue;
}
if (!FindSegmentation(target_text, word_res)) {
tprintf("APPLY_BOX: FAILURE: can't find segmentation for '%s'\n",
word->text());
pr_it.DeleteCurrentWord();
continue;
}
}
}
int
SOP_PrimGroupCentroid::bindToCentroids(fpreal t, int mode, int method)
{
int behavior;
exint int_value;
const GA_PrimitiveGroup *group;
GA_PrimitiveGroup *all_prims, *temp_group;
GA_Range pr_range;
GA_ROAttributeRef attr_gah, primattr_gah;
GA_ROHandleI class_h;
GA_ROHandleS str_h;
const GU_Detail *input_geo;
UT_Matrix4 mat;
UT_String attr_name, pattern, str_value;
UT_Vector3 pos;
// Get the second input geometry as read only.
GU_DetailHandleAutoReadLock gdl(inputGeoHandle(1));
input_geo = gdl.getGdp();
// Get the unmatched geometry behavior.
behavior = BEHAVIOR(t);
// Create a new attribute reference map.
GA_AttributeRefMap hmap(*gdp, input_geo);
// Get the attribute selection string.
BIND(pattern, t);
// If we have a pattern, try to build the ref map.
if (pattern.length() > 0)
buildRefMap(hmap, pattern, gdp, input_geo, mode, GA_ATTRIB_POINT);
// The list of GA_Primitives in the input geometry.
const GA_PrimitiveList &prim_list = gdp->getPrimitiveList();
// Create a temporary primitive group so we can keep track of all the
// primitives we have modified.
all_prims = createAdhocPrimGroup(*gdp, "allprims");
// Determine which attribute we need from the points, based on the mode.
switch (mode)
{
case 0:
attr_name = "group";
break;
case 1:
attr_name = "name";
break;
case 2:
attr_name = "class";
break;
default:
addError(SOP_MESSAGE, "Invalid mode setting");
return 1;
}
// Find the attribute.
attr_gah = input_geo->findPointAttribute(attr_name);
// If there is no attribute, add an error message and quit.
if (attr_gah.isInvalid())
{
addError(SOP_ATTRIBUTE_INVALID, attr_name);
return 1;
}
// If not using groups, we need to check if the matching primitive
// attribute exists on the geometry.
if (mode != 0)
{
// Try to find the attribute.
primattr_gah = gdp->findPrimitiveAttribute(attr_name);
// If there is no attribute, add an error message and quit.
if (primattr_gah.isInvalid())
{
addError(SOP_ATTRIBUTE_INVALID, attr_name);
return 1;
}
}
// 'class' uses the int handle.
if (mode == 2)
class_h.bind(attr_gah.getAttribute());
// Groups and 'name' use the string handle.
else
str_h.bind(attr_gah.getAttribute());
for (GA_Iterator it(input_geo->getPointRange()); !it.atEnd(); ++it)
{
if (mode == 0)
{
// Get the unique string value.
str_value = str_h.get(*it);
// Find the group on the geometry to bind.
group = gdp->findPrimitiveGroup(str_value);
// Ignore non-existent groups.
if (!group)
continue;
// Skip emptry groups.
if (group->isEmpty())
continue;
// The primtives in the group.
pr_range = gdp->getPrimitiveRange(group);
}
else
{
if (mode == 1)
{
// Get the unique string value.
str_value = str_h.get(*it);
// Get the prims with that string value.
pr_range = gdp->getRangeByValue(primattr_gah, str_value);
}
else
{
// Get the unique integer value.
int_value = class_h.get(*it);
// Get the prims with that integery value.
pr_range = gdp->getRangeByValue(primattr_gah, int_value);
}
// Create an adhoc group.
temp_group = createAdhocPrimGroup(*gdp);
temp_group->addRange(pr_range);
}
// Add the primitives in the range to the groups.
all_prims->addRange(pr_range);
// Bounding Box
if (method == 1)
{
// Calculate the bouding box center for this range.
boundingBox(gdp, pr_range, prim_list, pos);
}
// Center of Mass
else if (method == 2)
{
// Calculate the center of mass for this attribute value.
centerOfMass(pr_range, prim_list, pos);
}
// Barycenter
else
{
// Calculate the barycenter for this attribute value.
baryCenter(gdp, pr_range, prim_list, pos);
}
// Build the transform from the point information.
buildTransform(mat, input_geo, pos, *it);
// Transform the geometry from the centroid.
if (mode == 0)
gdp->transform(mat, group);
else
gdp->transform(mat, temp_group);
// Copy any necessary attributes from the incoming points to the
// geometry.
if (hmap.entries())
{
for (GA_Iterator pr_it(pr_range); !pr_it.atEnd(); ++pr_it)
{
hmap.copyValue(GA_ATTRIB_PRIMITIVE,
*pr_it,
GA_ATTRIB_POINT,
*it);
}
}
}
// We want to destroy prims that didn't have a matching name/group.
if (behavior)
{
// Flip the membership of all the prims that we did see.
all_prims->toggleEntries();
// Destroy the ones that we didn't.
gdp->deletePrimitives(*all_prims, true);
}
return 0;
}
OP_ERROR
SOP_Rain::cookMySop(OP_Context &context)
{
//UT_Interrupt *boss;
if (error() < UT_ERROR_ABORT)
{
//boss = UTgetInterrupt();
//boss->opStart("Start generating rain");
fpreal now = TIME(context.getTime());
long nPoints = NPOINTS( now );
UT_Vector3 rainDirection = RAINDIRECTION(now);
//rainDirection.normalize(); //TODO: check for (0,0,0) vector
RainData rain( now,
nPoints, BOUNDMIN (now), BOUNDMAX (now),
rainDirection,
DICEMIN(now), DICEMAX(now), SEED(now),
SPEED (now),
SPEEDVARIENCE (now));
if(rain.getAllocationState() == false || isPointsNumberChanged_ == true)
{
rain.allocate(nPoints);
}
if( rain.getAllocationState() == true &&
( rain.getCachedState() == false || isParameterChanged_ == true ) )
{
rain.computeInitialPositions();
rain.setCachedState(true);
}
if (isPointsGenerated_ == false)
{
printf("Generate Points procedure\n");
gdp->clearAndDestroy();
generatePoints(gdp, nPoints);
isPointsGenerated_ = true;
}
for ( GA_Iterator pr_it(gdp->getPrimitiveRange());
!pr_it.atEnd();
++pr_it)
{
GEO_Primitive* prim = gdp->getGEOPrimitive(*pr_it);
GA_Range range = prim->getPointRange();
rain.shiftPositions( gdp, range);
}
//boss->opEnd();
}
isParameterChanged_ = false;
isPointsNumberChanged_ = false;
//unlockInputs();
return error();
}
