Alignment-Lab-AI's picture
Upload folder using huggingface_hub
d5bfab8 verified
//! Detect grid patterns in images.
use super::{Histogram, Image, ImageDrawRect, ImageHistogram, ImageRotate90, Rectangle};
use std::collections::HashSet;
// ARC tasks where the grid cannot be detected
// 83302e8f - missing grid_mask for train3.input.
// 97239e3d - weird grid_mask for test1.input. It's good for the training pairs.
#[derive(Clone, Debug)]
pub struct GridPattern {
pub color: u8,
#[allow(dead_code)]
pub line_mask: Image,
#[allow(dead_code)]
pub intersection: u32,
#[allow(dead_code)]
pub union: u32,
#[allow(dead_code)]
pub jaccard_index: f32,
#[allow(dead_code)]
pub horizontal_line_count: u8,
#[allow(dead_code)]
pub horizontal_cell_count: u8,
#[allow(dead_code)]
pub vertical_line_count: u8,
#[allow(dead_code)]
pub vertical_cell_count: u8,
// Ideas for more:
// horizontal/vertical periodicity
// enumerated cell objects
// corner mask
// detect dots that makes up a grid
// convert pixel x,y coordinate to cell x,y coordinate
}
#[derive(Clone, Debug, PartialEq)]
struct Candidate {
color: u8,
combo: Combo,
#[allow(dead_code)]
combo_status: ComboStatus,
}
#[allow(dead_code)]
#[derive(Clone, Debug)]
pub struct Grid {
horizontal_candidates_full: Vec<Candidate>,
vertical_candidates_full: Vec<Candidate>,
horizontal_candidates_partial: Vec<Candidate>,
vertical_candidates_partial: Vec<Candidate>,
/// Disallow patterns with the same color.
///
/// The patterns are supposed to have unique colors, otherwise sorting the patterns would be non-deterministic.
patterns_full: Vec<GridPattern>,
patterns_partial: Vec<GridPattern>,
grid_found: bool,
grid_color: u8,
grid_with_mismatches_found: bool,
}
impl Grid {
pub fn analyze(image: &Image) -> anyhow::Result<Self> {
let mut instance = Self::new();
instance.perform_analyze(image)?;
// Future experiment:
// detect horizontal stacks. N cells wide, 1 cell tall.
// detect vertical stacks. 1 cell wide, N cells tall.
// enumerate cells
Ok(instance)
}
/// Is there an uninterrupted grid.
///
/// When a single color is used for both horizontal lines and vertical lines.
/// And the lines spans from edge to edge.
///
/// Then there is a grid. And this function returns `true`.
///
/// If it's interrupted, then it's more uncertain if there is a grid.
/// Here the function returns `false`.
pub fn grid_found(&self) -> bool {
self.grid_found
}
/// The color used for the uninterrupted grid.
/// Then there can only be a single color for the grid.
///
/// It's not possible to have an uninterrupted grid using multiple colors.
pub fn grid_color(&self) -> u8 {
self.grid_color
}
pub fn patterns_full(&self) -> &Vec<GridPattern> {
&self.patterns_full
}
#[allow(dead_code)]
pub fn patterns_partial(&self) -> &Vec<GridPattern> {
&self.patterns_partial
}
pub fn find_full_pattern_with_color(&self, color: u8) -> Option<&GridPattern> {
for pattern in &self.patterns_full {
if pattern.color == color {
return Some(pattern);
}
}
None
}
pub fn find_partial_pattern_with_color(&self, color: u8) -> Option<&GridPattern> {
for pattern in &self.patterns_partial {
if pattern.color == color {
return Some(pattern);
}
}
None
}
/// Is there a grid structure with a few mismatches.
///
/// This makes no sense for tiny images, smaller than 5 pixels.
///
/// The majority of pixels must agree on a single grid color.
///
/// The number of allowed mismatches depends on the size of the images.
/// - For bigger images, several mismatches are allowed.
/// - For medium sized images, fewer mismatches are allowed.
pub fn grid_with_mismatches_found(&self) -> bool {
self.grid_with_mismatches_found
}
fn new() -> Self {
Self {
horizontal_candidates_full: vec!(),
vertical_candidates_full: vec!(),
horizontal_candidates_partial: vec!(),
vertical_candidates_partial: vec!(),
patterns_full: vec!(),
patterns_partial: vec!(),
grid_found: false,
grid_color: u8::MAX,
grid_with_mismatches_found: false,
}
}
fn perform_analyze(&mut self, image: &Image) -> anyhow::Result<()> {
if image.width() < 2 || image.height() < 2 {
// Image is too small. Must be 2x2 or bigger
return Ok(());
}
let histogram: Histogram = image.histogram_all();
let unique_colors: u16 = histogram.number_of_counters_greater_than_zero();
if unique_colors < 2 {
// Too few colors to draw a grid
return Ok(());
}
self.perform_analyze_with_multiple_colors(image, true)?;
let rotated_image: Image = image.rotate_cw()?;
self.perform_analyze_with_multiple_colors(&rotated_image, false)?;
// println!("horizontal_candidates: {:?}", self.horizontal_candidates);
// println!("vertical_candidates: {:?}", self.vertical_candidates);
self.update_patterns(
image,
self.horizontal_candidates_full.clone(),
self.vertical_candidates_full.clone(),
true,
)?;
self.update_patterns(
image,
self.horizontal_candidates_partial.clone(),
self.vertical_candidates_partial.clone(),
false,
)?;
// The patterns are supposed to have unique colors.
// Thus sorting by color should be deterministic.
self.patterns_full.sort_unstable_by_key(|k| k.color);
self.patterns_partial.sort_unstable_by_key(|k| k.color);
Ok(())
}
fn update_patterns(&mut self, image: &Image, horizontal_candidates: Vec<Candidate>, vertical_candidates: Vec<Candidate>, is_full: bool) -> anyhow::Result<()> {
let mut candidate_colors = Histogram::new();
for candidate in &horizontal_candidates {
candidate_colors.increment(candidate.color);
}
for candidate in &vertical_candidates {
candidate_colors.increment(candidate.color);
}
let mut grid_found = false;
let mut grid_color = u8::MAX;
let mut grid_with_mismatches_found = false;
for (_count, color) in candidate_colors.pairs_descending() {
let candidate0: Option<&Candidate> = horizontal_candidates.iter().find(|candidate| candidate.color == color);
let candidate1: Option<&Candidate> = vertical_candidates.iter().find(|candidate| candidate.color == color);
let mut mask = Image::zero(image.width(), image.height());
let mut horizontal_lines = false;
let mut vertical_lines = false;
let mut horizontal_line_count: u8 = 0;
let mut horizontal_cell_count: u8 = 1;
let mut vertical_line_count: u8 = 0;
let mut vertical_cell_count: u8 = 1;
if let Some(candidate) = candidate1 {
(horizontal_line_count, horizontal_cell_count) = Self::draw_columns(&mut mask, candidate)?;
if candidate.combo_status.line_incorrect == 0 && candidate.combo_status.cell_incorrect == 0 {
horizontal_lines = true;
}
}
if let Some(candidate) = candidate0 {
(vertical_line_count, vertical_cell_count) = Self::draw_rows(&mut mask, candidate)?;
if candidate.combo_status.line_incorrect == 0 && candidate.combo_status.cell_incorrect == 0 {
vertical_lines = true;
}
}
let cell_count: u16 = (horizontal_cell_count as u16) * (vertical_cell_count as u16);
if cell_count < 2 {
continue;
}
let overlap_histogram: Histogram = image.histogram_with_mask(&mask)?;
let intersection: u32 = overlap_histogram.get(color);
let union: u32 = overlap_histogram.sum();
// println!("is_full: {} intersection: {} union: {}", is_full, intersection, union);
if intersection == 0 || union == 0 {
continue;
}
// println!("horizontal_lines: {}", horizontal_lines);
// println!("vertical_lines: {}", vertical_lines);
let jaccard_index: f32 = (intersection as f32) / (union as f32);
// println!("jaccard_index: {}", jaccard_index);
if is_full == false && jaccard_index < 0.5 {
// The grid pattern has low similarity with the image, ignoring.
// println!("ignoring grid that has low similarity with the image");
continue;
}
// println!("color: {} mask: {:?}", color, mask);
let pattern = GridPattern {
color,
line_mask: mask,
intersection,
union,
jaccard_index,
horizontal_line_count,
horizontal_cell_count,
vertical_line_count,
vertical_cell_count,
};
if is_full {
self.patterns_full.push(pattern);
} else {
self.patterns_partial.push(pattern);
}
if horizontal_lines && vertical_lines {
if intersection == union {
grid_found = true;
grid_color = color;
} else {
grid_with_mismatches_found = true;
}
}
}
// Only update grid_found when processing "full grid". Don't update grid_found when analyzing partial patterns
if is_full {
self.grid_found = grid_found;
self.grid_color = grid_color;
}
// Only update when processing "partial grid patterns". Don't update when processing "full grid".
if !is_full {
self.grid_with_mismatches_found = grid_with_mismatches_found;
}
Ok(())
}
fn draw_columns(result_image: &mut Image, candidate: &Candidate) -> anyhow::Result<(u8, u8)> {
let mut x: i16 = candidate.combo.initial_position;
let width: i16 = result_image.width() as i16;
let mut mask: Image = result_image.clone();
let mut line_count: u8 = 0;
let mut cell_count: u8 = 0;
'outer: for _ in 0..30 {
let mut line_count_increment: u8 = 1;
for _ in 0..candidate.combo.line_size {
if x >= 0 && x < width {
line_count += line_count_increment;
line_count_increment = 0;
let xx = (x & 255) as u8;
let r = Rectangle::new(xx, 0, 1, result_image.height());
mask = mask.draw_rect_filled(r, 1)?;
}
x += 1;
if x >= width {
break 'outer;
}
}
let mut cell_count_increment: u8 = 1;
for _ in 0..candidate.combo.cell_size {
if x >= 0 && x < width {
cell_count += cell_count_increment;
cell_count_increment = 0;
}
x += 1;
if x >= width {
break 'outer;
}
}
}
result_image.set_image(mask);
Ok((line_count, cell_count))
}
fn draw_rows(result_image: &mut Image, candidate: &Candidate) -> anyhow::Result<(u8, u8)> {
let mut y: i16 = candidate.combo.initial_position;
let height: i16 = result_image.height() as i16;
let mut mask: Image = result_image.clone();
let mut line_count: u8 = 0;
let mut cell_count: u8 = 0;
'outer: for _ in 0..30 {
let mut line_count_increment: u8 = 1;
for _ in 0..candidate.combo.line_size {
if y >= 0 && y < height {
line_count += line_count_increment;
line_count_increment = 0;
let yy = (y & 255) as u8;
let r = Rectangle::new(0, yy, result_image.width(), 1);
mask = mask.draw_rect_filled(r, 1)?;
}
y += 1;
if y >= height {
break 'outer;
}
}
let mut cell_count_increment: u8 = 1;
for _ in 0..candidate.combo.cell_size {
if y >= 0 && y < height {
cell_count += cell_count_increment;
cell_count_increment = 0;
}
y += 1;
if y >= height {
break 'outer;
}
}
}
result_image.set_image(mask);
Ok((line_count, cell_count))
}
fn perform_analyze_with_multiple_colors(&mut self, image: &Image, is_horizontal: bool) -> anyhow::Result<()> {
let rows: Vec<Histogram> = image.histogram_rows();
let mut full_row_colors = Vec::<Option<u8>>::new();
let mut partial_row_colors = Vec::<Option<u8>>::new();
let mut rows_histogram = Histogram::new();
for (_y, histogram) in rows.iter().enumerate() {
let unique_colors: u16 = histogram.number_of_counters_greater_than_zero();
if unique_colors != 1 && image.width() < 5 {
full_row_colors.push(None);
partial_row_colors.push(None);
continue;
}
let (color, count) = match histogram.most_popular_pair_disallow_ambiguous() {
Some(value) => value,
None => {
full_row_colors.push(None);
partial_row_colors.push(None);
continue;
}
};
// Detect grid and allow for some mismatches.
// It's a full line when it spans from edge to edge uninterrupted with just one color.
// It's a partial line when it's interrupted and contains a few pixels with a different color.
if count < (image.width() as u32) * 7 / 10 {
full_row_colors.push(None);
partial_row_colors.push(None);
continue;
}
// println!("row y: {} color: {}", y, color);
if count == image.width() as u32 {
full_row_colors.push(Some(color));
partial_row_colors.push(Some(color));
rows_histogram.increment(color);
} else {
full_row_colors.push(None);
partial_row_colors.push(Some(color));
rows_histogram.increment(color);
}
}
// Process full lines
// measure spacing between the lines, thickness of lines
let mut full_candidates = Vec::<Candidate>::new();
for (_count, color) in rows_histogram.pairs_descending() {
let (combo, combo_status) = match Self::measure(color, &full_row_colors) {
Ok(value) => value,
_ => continue
};
let candidate = Candidate {
color,
combo,
combo_status,
};
full_candidates.push(candidate);
}
// Process partial+full lines
// When a partial candidate is identical to an already found full_candidate, then discard the partial candidate
let mut partial_candidates = Vec::<Candidate>::new();
for (_count, color) in rows_histogram.pairs_descending() {
let (combo, combo_status) = match Self::measure(color, &partial_row_colors) {
Ok(value) => value,
_ => continue
};
let candidate = Candidate {
color,
combo,
combo_status,
};
if full_candidates.contains(&candidate) {
// println!("partial candidate is identical to already found full candidate. Ignoring the partial candidate");
continue;
}
partial_candidates.push(candidate);
}
if is_horizontal {
self.horizontal_candidates_full = full_candidates;
self.horizontal_candidates_partial = partial_candidates;
} else {
self.vertical_candidates_full = full_candidates;
self.vertical_candidates_partial = partial_candidates;
}
Ok(())
}
fn measure(measure_color: u8, row_colors: &Vec<Option<u8>>) -> anyhow::Result<(Combo, ComboStatus)> {
let mut found_max_possible_line_size: u8 = 0;
let mut current_possible_line_size: u8 = 0;
let mut found_max_possible_cell_size: u8 = 0;
let mut current_possible_cell_size: u8 = 0;
let mut positions = Vec::<u8>::new();
let mut position_set = HashSet::<i16>::new();
for (index, row_color) in row_colors.iter().enumerate() {
if *row_color != Some(measure_color) {
current_possible_line_size = 0;
if current_possible_cell_size < u8::MAX {
current_possible_cell_size += 1;
}
if current_possible_cell_size > found_max_possible_cell_size {
found_max_possible_cell_size = current_possible_cell_size;
}
continue;
}
current_possible_cell_size = 0;
let position: u8 = (index & 255) as u8;
positions.push(position);
position_set.insert(position as i16);
if current_possible_line_size < u8::MAX {
current_possible_line_size += 1;
}
if current_possible_line_size > found_max_possible_line_size {
found_max_possible_line_size = current_possible_line_size;
}
}
if positions.is_empty() {
return Err(anyhow::anyhow!("positions. Found none"));
}
if found_max_possible_line_size == 0 {
return Err(anyhow::anyhow!("found_max_possible_line_size"));
}
if found_max_possible_cell_size == 0 {
return Err(anyhow::anyhow!("found_max_possible_cell_size"));
}
let max_line_size: u8 = found_max_possible_line_size;
let max_cell_size: u8 = found_max_possible_cell_size;
// println!("color: {} positions: {:?}", color, positions);
// println!("max_line_size: {}", max_line_size);
// println!("max_cell_size: {}", max_cell_size);
let mut best = ComboStatus {
line_correct: 0,
line_incorrect: u8::MAX,
cell_correct: 0,
cell_incorrect: u8::MAX
};
let mut current_error: i32 = i32::MIN;
let mut found_combo: Option<Combo> = None;
let max_position: i16 = ((row_colors.len() & 255) as i16) - 1;
for cell_size in 1..=max_cell_size {
for line_size in 1..=max_line_size {
let periodicity_u16: u16 = (cell_size as u16) + (line_size as u16);
let periodicity: u8 = (periodicity_u16 & 255) as u8;
for offset in 0..periodicity {
let initial_position: i16 = -(offset as i16);
let combo = Combo {
initial_position,
line_size,
cell_size
};
let status: ComboStatus = combo.score(max_position, &position_set);
let error: i32 = status.error();
if error > current_error {
current_error = error;
best = status;
found_combo = Some(combo);
}
}
}
}
// pick combo with optimal score
let combo: Combo = match found_combo {
Some(value) => value,
None => {
return Err(anyhow::anyhow!("unable to find a combo that fits the data"));
}
};
// println!("found combo: {:?} status: {:?} error: {}", combo, best, current_error);
Ok((combo, best))
}
}
#[derive(Clone, Debug, PartialEq)]
struct Combo {
initial_position: i16,
line_size: u8,
cell_size: u8
}
#[derive(Clone, Debug, PartialEq)]
struct ComboStatus {
line_correct: u8,
line_incorrect: u8,
cell_correct: u8,
cell_incorrect: u8,
}
impl ComboStatus {
fn error(&self) -> i32 {
let line_correct2: u16 = (self.line_correct as u16) * (self.line_correct as u16);
let cell_correct2: u16 = (self.cell_correct as u16) * (self.cell_correct as u16);
let line_incorrect2: u16 = (self.line_incorrect as u16) * (self.line_incorrect as u16);
let cell_incorrect2: u16 = (self.cell_incorrect as u16) * (self.cell_incorrect as u16);
let sum: i32 = (line_correct2 as i32) + (cell_correct2 as i32) - (line_incorrect2 as i32) - (cell_incorrect2 as i32);
sum
}
}
impl Combo {
fn score(&self, max_position: i16, position_set: &HashSet<i16>) -> ComboStatus {
let mut line_correct: u8 = 0;
let mut line_incorrect: u8 = 0;
let mut cell_correct: u8 = 0;
let mut cell_incorrect: u8 = 0;
let mut current_position: i16 = self.initial_position;
let biggest_arc_grid_size: u8 = 30 * 2;
for _ in 0..biggest_arc_grid_size {
for _ in 0..self.line_size {
if current_position >= 0 && current_position <= max_position {
if position_set.contains(&current_position) {
line_correct += 1;
} else {
line_incorrect += 1;
}
}
current_position += 1;
}
for _ in 0..self.cell_size {
if current_position >= 0 && current_position <= max_position {
if !position_set.contains(&current_position) {
cell_correct += 1;
} else {
cell_incorrect += 1;
}
}
current_position += 1;
}
if current_position > max_position {
break;
}
}
let status = ComboStatus {
line_correct,
line_incorrect,
cell_correct,
cell_incorrect,
};
// println!("score: {} {} {} -> {} {} {} {}", self.initial_position, self.line_size, self.cell_size, line_correct, line_incorrect, cell_correct, cell_incorrect);
status
}
}
#[cfg(test)]
mod tests {
use super::*;
use crate::arc::ImageTryCreate;
#[test]
fn test_10000_gridsize1_cellsize1() {
// Arrange
let pixels: Vec<u8> = vec![
9, 9, 9, 9, 9,
9, 7, 9, 7, 9,
9, 9, 9, 9, 9,
9, 7, 9, 7, 9,
9, 9, 9, 9, 9,
];
let input: Image = Image::try_create(5, 5, pixels).expect("image");
// Act
let instance = Grid::analyze(&input).expect("ok");
// Assert
let pattern: &GridPattern = instance.patterns_full.first().expect("GridPattern");
let expected_pixels: Vec<u8> = vec![
1, 1, 1, 1, 1,
1, 0, 1, 0, 1,
1, 1, 1, 1, 1,
1, 0, 1, 0, 1,
1, 1, 1, 1, 1,
];
let expected: Image = Image::try_create(5, 5, expected_pixels).expect("image");
assert_eq!(pattern.line_mask, expected);
}
#[test]
fn test_10001_gridsize1_cellsize3() {
// Arrange
let pixels: Vec<u8> = vec![
5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
5, 9, 9, 9, 5, 9, 9, 9, 5, 9, 9, 9, 5,
5, 9, 9, 9, 5, 9, 9, 9, 5, 9, 9, 9, 5,
5, 9, 9, 9, 5, 9, 9, 9, 5, 9, 9, 9, 5,
5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
5, 9, 9, 9, 5, 9, 9, 9, 5, 9, 9, 9, 5,
5, 9, 9, 9, 5, 9, 9, 9, 5, 9, 9, 9, 5,
5, 9, 9, 9, 5, 9, 9, 9, 5, 9, 9, 9, 5,
5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
];
let input: Image = Image::try_create(13, 9, pixels).expect("image");
// Act
let instance = Grid::analyze(&input).expect("ok");
// Assert
let pattern: &GridPattern = instance.patterns_full.first().expect("GridPattern");
let expected_pixels: Vec<u8> = vec![
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1,
1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1,
1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1,
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1,
1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1,
1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1,
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
];
let expected: Image = Image::try_create(13, 9, expected_pixels).expect("image");
assert_eq!(pattern.line_mask, expected);
}
#[test]
fn test_10002_gridsize2_cellsize1() {
// Arrange
let pixels: Vec<u8> = vec![
7, 7, 7, 7, 7, 7, 7, 7,
7, 7, 7, 7, 7, 7, 7, 7,
7, 7, 9, 7, 7, 9, 7, 7,
7, 7, 7, 7, 7, 7, 7, 7,
7, 7, 7, 7, 7, 7, 7, 7,
7, 7, 9, 7, 7, 9, 7, 7,
7, 7, 7, 7, 7, 7, 7, 7,
7, 7, 7, 7, 7, 7, 7, 7,
];
let input: Image = Image::try_create(8, 8, pixels).expect("image");
// Act
let instance = Grid::analyze(&input).expect("ok");
// Assert
let pattern: &GridPattern = instance.patterns_full.first().expect("GridPattern");
let expected_pixels: Vec<u8> = vec![
1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 0, 1, 1, 0, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 0, 1, 1, 0, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1,
];
let expected: Image = Image::try_create(8, 8, expected_pixels).expect("image");
assert_eq!(pattern.line_mask, expected);
}
#[test]
fn test_10003_gridsize3_offset2_cellsize1() {
// Arrange
let pixels: Vec<u8> = vec![
0, 0, 0, 0, 0, 0, 0,
0, 1, 0, 0, 0, 1, 0,
0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0,
0, 1, 0, 0, 0, 1, 0,
0, 0, 0, 0, 0, 0, 0,
];
let input: Image = Image::try_create(7, 7, pixels).expect("image");
// Act
let instance = Grid::analyze(&input).expect("ok");
// Assert
let pattern: &GridPattern = instance.patterns_full.first().expect("GridPattern");
let expected_pixels: Vec<u8> = vec![
1, 1, 1, 1, 1, 1, 1,
1, 0, 1, 1, 1, 0, 1,
1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1,
1, 0, 1, 1, 1, 0, 1,
1, 1, 1, 1, 1, 1, 1,
];
let expected: Image = Image::try_create(7, 7, expected_pixels).expect("image");
assert_eq!(pattern.line_mask, expected);
}
#[test]
fn test_10004_two_grids_with_different_size() {
// Arrange
let pixels: Vec<u8> = vec![
1, 1, 1, 1,
2, 2, 2, 2,
0, 0, 0, 0,
0, 0, 0, 0,
1, 1, 1, 1,
0, 0, 0, 0,
2, 2, 2, 2,
0, 0, 0, 0,
1, 1, 1, 1,
0, 0, 0, 0,
0, 0, 0, 0,
2, 2, 2, 2,
1, 1, 1, 1,
0, 0, 0, 0,
];
let input: Image = Image::try_create(4, 14, pixels).expect("image");
// Act
let instance = Grid::analyze(&input).expect("ok");
// Assert
assert_eq!(instance.patterns_full.len(), 3);
{
let pattern: &GridPattern = &instance.patterns_full[1];
let expected_pixels: Vec<u8> = vec![
1, 1, 1, 1,
0, 0, 0, 0,
0, 0, 0, 0,
0, 0, 0, 0,
1, 1, 1, 1,
0, 0, 0, 0,
0, 0, 0, 0,
0, 0, 0, 0,
1, 1, 1, 1,
0, 0, 0, 0,
0, 0, 0, 0,
0, 0, 0, 0,
1, 1, 1, 1,
0, 0, 0, 0,
];
let expected: Image = Image::try_create(4, 14, expected_pixels).expect("image");
assert_eq!(pattern.line_mask, expected);
}
{
let pattern: &GridPattern = &instance.patterns_full[2];
let expected_pixels: Vec<u8> = vec![
0, 0, 0, 0,
1, 1, 1, 1,
0, 0, 0, 0,
0, 0, 0, 0,
0, 0, 0, 0,
0, 0, 0, 0,
1, 1, 1, 1,
0, 0, 0, 0,
0, 0, 0, 0,
0, 0, 0, 0,
0, 0, 0, 0,
1, 1, 1, 1,
0, 0, 0, 0,
0, 0, 0, 0,
];
let expected: Image = Image::try_create(4, 14, expected_pixels).expect("image");
assert_eq!(pattern.line_mask, expected);
}
}
#[test]
fn test_10005_split_middle() {
// Arrange
let pixels: Vec<u8> = vec![
9, 9, 9, 7, 7, 9, 9, 9,
9, 9, 9, 7, 7, 9, 9, 9,
7, 7, 7, 7, 7, 7, 7, 7,
9, 9, 9, 7, 7, 9, 9, 9,
9, 9, 9, 7, 7, 9, 9, 9,
];
let input: Image = Image::try_create(8, 5, pixels).expect("image");
// Act
let instance = Grid::analyze(&input).expect("ok");
// Assert
let pattern: &GridPattern = instance.patterns_full.first().expect("GridPattern");
let expected_pixels: Vec<u8> = vec![
0, 0, 0, 1, 1, 0, 0, 0,
0, 0, 0, 1, 1, 0, 0, 0,
1, 1, 1, 1, 1, 1, 1, 1,
0, 0, 0, 1, 1, 0, 0, 0,
0, 0, 0, 1, 1, 0, 0, 0,
];
let expected: Image = Image::try_create(8, 5, expected_pixels).expect("image");
assert_eq!(pattern.line_mask, expected);
}
#[test]
fn test_10006_detect_grid() {
// Arrange
let pixels: Vec<u8> = vec![
0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 5, 0, 0, 0,
0, 0, 0, 5, 0, 0, 0,
0, 5, 0, 5, 0, 0, 0,
0, 5, 0, 5, 0, 0, 0,
0, 5, 0, 5, 0, 5, 0,
0, 5, 0, 5, 0, 5, 0,
];
let input: Image = Image::try_create(7, 7, pixels).expect("image");
// Act
let instance = Grid::analyze(&input).expect("ok");
// Assert
let pattern: &GridPattern = instance.patterns_full.first().expect("GridPattern");
let expected_pixels: Vec<u8> = vec![
1, 1, 1, 1, 1, 1, 1,
1, 0, 1, 0, 1, 0, 1,
1, 0, 1, 0, 1, 0, 1,
1, 0, 1, 0, 1, 0, 1,
1, 0, 1, 0, 1, 0, 1,
1, 0, 1, 0, 1, 0, 1,
1, 0, 1, 0, 1, 0, 1,
];
let expected: Image = Image::try_create(7, 7, expected_pixels).expect("image");
assert_eq!(pattern.line_mask, expected);
}
#[test]
fn test_10007_detect_grid() {
// Arrange
let pixels: Vec<u8> = vec![
0, 0, 0, 0, 0, 0, 0,
0, 8, 0, 0, 0, 0, 0,
0, 8, 8, 0, 0, 0, 0,
0, 0, 0, 0, 8, 8, 0,
0, 0, 0, 0, 0, 8, 0,
0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0,
];
let input: Image = Image::try_create(7, 7, pixels).expect("image");
// Act
let instance = Grid::analyze(&input).expect("ok");
// Assert
let pattern: &GridPattern = instance.patterns_full.first().expect("GridPattern");
let expected_pixels: Vec<u8> = vec![
1, 1, 1, 1, 1, 1, 1,
1, 0, 0, 1, 0, 0, 1,
1, 0, 0, 1, 0, 0, 1,
1, 0, 0, 1, 0, 0, 1,
1, 0, 0, 1, 0, 0, 1,
1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1,
];
let expected: Image = Image::try_create(7, 7, expected_pixels).expect("image");
assert_eq!(pattern.line_mask, expected);
}
#[test]
fn test_10008_detect_grid() {
// Arrange
let pixels: Vec<u8> = vec![
1, 1, 0, 0, 1,
0, 0, 0, 0, 1,
0, 0, 0, 0, 0,
0, 0, 0, 2, 2,
1, 1, 0, 2, 2,
];
let input: Image = Image::try_create(5, 5, pixels).expect("image");
// Act
let instance = Grid::analyze(&input).expect("ok");
// Assert
assert_eq!(instance.grid_found, true);
assert_eq!(instance.grid_color, 0);
assert_eq!(instance.patterns_full.len(), 1);
let pattern: &GridPattern = instance.patterns_full.first().expect("GridPattern");
let expected_pixels: Vec<u8> = vec![
0, 0, 1, 0, 0,
0, 0, 1, 0, 0,
1, 1, 1, 1, 1,
0, 0, 1, 0, 0,
0, 0, 1, 0, 0,
];
let expected: Image = Image::try_create(5, 5, expected_pixels).expect("image");
assert_eq!(pattern.line_mask, expected);
}
#[test]
fn test_20000_partial_grid() {
// Arrange
let pixels: Vec<u8> = vec![
1, 1, 5, 2, 2,
1, 1, 5, 2, 2,
5, 5, 7, 5, 5,
3, 3, 5, 4, 4,
3, 3, 5, 4, 4,
];
let input: Image = Image::try_create(5, 5, pixels).expect("image");
// Act
let instance = Grid::analyze(&input).expect("ok");
// Assert
assert_eq!(instance.grid_found(), false);
assert_eq!(instance.patterns_partial.len(), 1);
let pattern: &GridPattern = instance.patterns_partial.first().expect("GridPattern");
let expected_pixels: Vec<u8> = vec![
0, 0, 1, 0, 0,
0, 0, 1, 0, 0,
1, 1, 1, 1, 1,
0, 0, 1, 0, 0,
0, 0, 1, 0, 0,
];
let expected: Image = Image::try_create(5, 5, expected_pixels).expect("image");
assert_eq!(pattern.line_mask, expected);
assert_eq!(pattern.color, 5);
}
#[test]
fn test_20001_partial_grid() {
// Arrange
let pixels: Vec<u8> = vec![
2, 2, 5, 2, 2,
2, 2, 5, 2, 2,
5, 5, 2, 5, 5,
2, 2, 5, 2, 2,
2, 2, 5, 2, 2,
];
let input: Image = Image::try_create(5, 5, pixels).expect("image");
// Act
let instance = Grid::analyze(&input).expect("ok");
// Assert
assert_eq!(instance.grid_found(), false);
assert_eq!(instance.patterns_partial.len(), 1);
let pattern: &GridPattern = instance.patterns_partial.first().expect("GridPattern");
let expected_pixels: Vec<u8> = vec![
0, 0, 1, 0, 0,
0, 0, 1, 0, 0,
1, 1, 1, 1, 1,
0, 0, 1, 0, 0,
0, 0, 1, 0, 0,
];
let expected: Image = Image::try_create(5, 5, expected_pixels).expect("image");
assert_eq!(pattern.line_mask, expected);
assert_eq!(pattern.color, 5);
}
#[test]
fn test_20002_partial_grid() {
// Arrange
let pixels: Vec<u8> = vec![
2, 2, 5, 2, 2, 2, 5,
2, 2, 5, 2, 2, 2, 5,
5, 5, 2, 5, 5, 5, 2,
2, 2, 5, 2, 2, 2, 5,
2, 2, 5, 2, 2, 2, 5,
2, 2, 5, 2, 2, 2, 5,
5, 5, 2, 5, 5, 5, 2,
2, 2, 5, 2, 2, 2, 5,
];
let input: Image = Image::try_create(7, 8, pixels).expect("image");
// Act
let instance = Grid::analyze(&input).expect("ok");
// Assert
assert_eq!(instance.grid_found(), false);
assert_eq!(instance.patterns_partial.len(), 2);
let pattern: &GridPattern = instance.find_partial_pattern_with_color(5).expect("GridPattern");
let expected_pixels: Vec<u8> = vec![
0, 0, 1, 0, 0, 0, 1,
0, 0, 1, 0, 0, 0, 1,
1, 1, 1, 1, 1, 1, 1,
0, 0, 1, 0, 0, 0, 1,
0, 0, 1, 0, 0, 0, 1,
0, 0, 1, 0, 0, 0, 1,
1, 1, 1, 1, 1, 1, 1,
0, 0, 1, 0, 0, 0, 1,
];
let expected: Image = Image::try_create(7, 8, expected_pixels).expect("image");
assert_eq!(pattern.line_mask, expected);
assert_eq!(pattern.color, 5);
}
#[test]
fn test_20003_partial_grid() {
// Arrange
let pixels: Vec<u8> = vec![
0, 0, 0, 2, 5, 0, 0, 2, 0, 5, 0, 0, 0,
2, 0, 0, 0, 5, 0, 2, 0, 0, 5, 0, 0, 0,
0, 0, 0, 0, 2, 0, 2, 0, 0, 5, 0, 0, 0,
0, 0, 0, 0, 5, 0, 0, 0, 0, 5, 0, 0, 0,
2, 5, 5, 5, 5, 5, 5, 5, 2, 5, 5, 5, 5,
0, 2, 0, 0, 5, 0, 2, 0, 0, 5, 0, 0, 0,
0, 0, 2, 0, 5, 0, 0, 2, 0, 5, 0, 2, 0,
0, 0, 0, 0, 5, 0, 0, 0, 0, 5, 0, 0, 0,
0, 0, 0, 0, 5, 0, 0, 0, 0, 5, 0, 0, 0,
5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 2, 5, 5,
0, 0, 0, 0, 5, 0, 0, 0, 0, 5, 0, 0, 0,
0, 2, 0, 0, 5, 0, 2, 0, 0, 5, 0, 0, 0,
0, 0, 0, 0, 5, 0, 0, 0, 0, 5, 0, 0, 2,
0, 0, 0, 0, 5, 0, 0, 0, 0, 5, 0, 0, 0,
5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
0, 0, 0, 0, 5, 0, 0, 0, 0, 5, 0, 0, 0,
0, 0, 0, 0, 5, 0, 0, 0, 0, 5, 0, 0, 0,
0, 0, 0, 0, 5, 0, 0, 0, 0, 5, 0, 0, 0,
0, 0, 0, 0, 5, 0, 0, 0, 0, 5, 2, 0, 0,
];
let input: Image = Image::try_create(13, 19, pixels).expect("image");
// Act
let instance = Grid::analyze(&input).expect("ok");
// Assert
assert_eq!(instance.grid_found(), true);
assert_eq!(instance.patterns_partial.len(), 2);
let pattern: &GridPattern = instance.find_partial_pattern_with_color(5).expect("GridPattern");
let expected_pixels: Vec<u8> = vec![
0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0,
0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0,
0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0,
0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0,
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0,
0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0,
0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0,
0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0,
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0,
0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0,
0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0,
0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0,
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0,
0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0,
0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0,
0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0,
];
let expected: Image = Image::try_create(13, 19, expected_pixels).expect("image");
assert_eq!(pattern.line_mask, expected);
assert_eq!(pattern.color, 5);
}
}