import numpy as np import cv2 def crop_and_scaled_imgs(imgs): PAD = 5 # use the last image to find the bounding box of the non-white area and the transformation parameters # and then apply the transformation to all images img = imgs[-1] gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) # Threshold the image to create a binary mask _, binary_mask = cv2.threshold(gray, 240, 255, cv2.THRESH_BINARY_INV) # Find the coordinates of non-zero pixels coords = cv2.findNonZero(binary_mask) # Get the bounding box of the non-zero pixels x, y, w, h = cv2.boundingRect(coords) x = max(0, x-PAD) y = max(0, y-PAD) x_end = min(img.shape[1], x+w+2*PAD) y_end = min(img.shape[0], y+h+2*PAD) w = x_end - x h = y_end - y SIZE = 400 # Calculate the position to center the ROI in the SIZExSIZE image start_x = max(0, (SIZE - w) // 2) start_y = max(0, (SIZE - h) // 2) # Create a new SIZExSIZE rgb images new_imgs = [np.ones((SIZE, SIZE, 3), dtype=np.uint8) * 255 for _ in range(len(imgs))] for i in range(len(imgs)): # Extract the ROI (region of interest) of the non-white area roi = imgs[i][y:y+h, x:x+w] # If the ROI is larger than 256x256, resize it if w > SIZE or h > SIZE: scale = min(SIZE / w, SIZE / h) new_w = int(w * scale) new_h = int(h * scale) roi = cv2.resize(roi, (new_w, new_h), interpolation=cv2.INTER_AREA) else: new_w = w new_h = h # new_imgs[i] = np.ones((256, 256), dtype=np.uint8) * 255 # centered_img = np.ones((256, 256), dtype=np.uint8) * 255 # Place the ROI in the centered position new_imgs[i][start_y:start_y+new_h, start_x:start_x+new_w] = roi return new_imgs HALF_INF = 63 INF = 126 EPS_DIST = 1/20 EPS_ANGLE = 2.86 SCALE = 15 MOVE_SPEED = 25 ROTATE_SPEED = 30 FPS = 24 class Turtle: def __init__(self, canvas_size=(1000, 1000)): self.x = canvas_size[0] // 2 self.y = canvas_size[1] // 2 self.heading = 0 self.canvas = np.ones((canvas_size[1], canvas_size[0], 3), dtype=np.uint8) * 255 self.is_down = True self.time_since_last_frame = 0 self.frames = [self.canvas.copy()] def forward(self, dist): # print('st', self.x, self.y) # self.forward_step(dist * SCALE) # print('ed', self.x, self.y) # return dist = dist * SCALE sign = 1 if dist > 0 else -1 abs_dist = abs(dist) if self.time_since_last_frame + abs_dist / MOVE_SPEED >= 1: dist1 = (1 - self.time_since_last_frame) * MOVE_SPEED self.forward_step(dist1 * sign) self.save_frame_with_turtle() self.time_since_last_frame = 0 # for loop to step forward num_steps = int((abs_dist - dist1) / MOVE_SPEED) for _ in range(num_steps): self.forward_step(MOVE_SPEED * sign) self.save_frame_with_turtle() last_abs_dist = abs_dist - dist1 - num_steps * MOVE_SPEED if last_abs_dist >= MOVE_SPEED: self.forward_step(MOVE_SPEED * sign) self.save_frame_with_turtle() last_abs_dist -= MOVE_SPEED self.forward_step(last_abs_dist * sign) self.time_since_last_frame = last_abs_dist / MOVE_SPEED else: self.forward_step(abs_dist * sign) # self.time_since_last_frame += abs_dist / MOVE_SPEED # if self.time_since_last_frame >= 1: # self.time_since_last_frame = 0 def forward_step(self, dist): # print('step', dist) if dist == 0: return x0, y0 = self.x, self.y x1 = (x0 + dist * np.cos(self.heading)) y1 = (y0 - dist * np.sin(self.heading)) if self.is_down: cv2.line(self.canvas, (int(np.rint(x0)), int(np.rint(y0))), (int(np.rint(x1)), int(np.rint(y1))), (0, 0, 0), 3) self.x, self.y = x1, y1 self.time_since_last_frame += abs(dist) / MOVE_SPEED # self.frames.append(self.canvas.copy()) # self.save_frame_with_turtle() # print(self.x, self.y) def save_frame_with_turtle(self): # save the current frame to frames buffer # also plot a red triangle to represent the turtle pointing to the current direction # draw the turtle x, y = self.x, self.y canvas_copy = self.canvas.copy() triangle_size = 10 x0 = int(np.rint(x + triangle_size * np.cos(self.heading))) y0 = int(np.rint(y - triangle_size * np.sin(self.heading))) x1 = int(np.rint(x + triangle_size * np.cos(self.heading + 2 * np.pi / 3))) y1 = int(np.rint(y - triangle_size * np.sin(self.heading + 2 * np.pi / 3))) x2 = int(np.rint(x + triangle_size * np.cos(self.heading - 2 * np.pi / 3))) y2 = int(np.rint(y - triangle_size * np.sin(self.heading - 2 * np.pi / 3))) x3 = int(np.rint(x - 0.25 * triangle_size * np.cos(self.heading))) y3 = int(np.rint(y + 0.25 * triangle_size * np.sin(self.heading))) # fill the triangle cv2.fillPoly(canvas_copy, [np.array([(x0, y0), (x1, y1), (x3, y3), (x2, y2)], dtype=np.int32)], (0, 0, 255)) self.frames.append(canvas_copy) def left(self, angle): # print('angel', angle) # print('ast', self.heading) # self.heading += angle * np.pi / 180 self.turn_to(angle) # print('aed', self.heading) def right(self, angle): # print('angel', angle) # print('ast', self.heading) # self.heading -= angle * np.pi / 180 self.turn_to(-angle) # print('aed', self.heading) def turn_to(self, angle): abs_angle = abs(angle) sign = 1 if angle > 0 else -1 if self.time_since_last_frame + abs(angle) / ROTATE_SPEED > 1: angle1 = (1 - self.time_since_last_frame) * ROTATE_SPEED self.turn_to_step(angle1 * sign) self.save_frame_with_turtle() self.time_since_last_frame = 0 num_steps = int((abs_angle - angle1) / ROTATE_SPEED) for _ in range(num_steps): self.turn_to_step(ROTATE_SPEED * sign) self.save_frame_with_turtle() last_abs_angle = abs_angle - angle1 - num_steps * ROTATE_SPEED if last_abs_angle >= ROTATE_SPEED: self.turn_to_step(ROTATE_SPEED * sign) self.save_frame_with_turtle() last_abs_angle -= ROTATE_SPEED self.turn_to_step(last_abs_angle * sign) self.time_since_last_frame = last_abs_angle / ROTATE_SPEED else: self.turn_to_step(abs_angle * sign) # self.time_since_last_frame += abs_angle / ROTATE_SPEED def turn_to_step(self, angle): # print('turn step', angle) self.heading += angle * np.pi / 180 self.time_since_last_frame += abs(angle) / ROTATE_SPEED def penup(self): self.is_down = False def pendown(self): self.is_down = True def save(self, path): if path: cv2.imwrite(path, self.canvas) return self.canvas def save_gif(self, path): import imageio.v3 as iio frames_rgb = [cv2.cvtColor(frame, cv2.COLOR_BGR2RGB) for frame in self.frames] print(f'number of frames: {len(frames_rgb)}') frames_rgb.extend(FPS*2 * [frames_rgb[-1]]) frames_rgb = crop_and_scaled_imgs(frames_rgb) # iio.imwrite(path, np.stack(frames_rgb), fps=30, plugin='pillow') return iio.imwrite('', np.stack(frames_rgb), fps=FPS, loop=0, plugin='pillow', format='gif') class _TurtleState: def __init__(self, turtle): self.turtle = turtle self.position = None self.heading = None self.pen_status = None def __enter__(self): self.position = (self.turtle.x, self.turtle.y) self.heading = self.turtle.heading self.pen_status = self.turtle.is_down return self def __exit__(self, exc_type, exc_val, exc_tb): self.turtle.penup() self.turtle.x, self.turtle.y = self.position self.turtle.heading = self.heading if self.pen_status: self.turtle.pendown() if __name__ == "__main__": turtle = Turtle() def forward(dist): turtle.forward(dist) def left(angle): turtle.left(angle) def right(angle): turtle.right(angle) def penup(): turtle.penup() def pendown(): turtle.pendown() def save(path): turtle.save(path) def fork_state(): """ Clone the current state of the turtle. Usage: with clone_state(): forward(100) left(90) forward(100) """ return turtle._TurtleState(turtle) # Example usage def example_plot(): forward(5) with fork_state(): forward(10) left(90) forward(10) with fork_state(): right(90) forward(20) left(90) forward(10) left(90) forward(10) right(90) forward(50) save("test2.png") return turtle.frames def plot2(): for j in range(2): forward(2) left(0.0) for i in range(4): forward(2) left(90) forward(0) left(180.0) forward(2) left(180.0) FINAL_IMAGE = turtle.save("") def plot3(): frames = [] frames.append(np.array(turtle.save(""))) for j in range(2): forward(2) frames.append(np.array(turtle.save(""))) left(0.0) for i in range(4): forward(2) left(90) frames.append(np.array(turtle.save(""))) forward(0) left(180.0) forward(2) left(180.0) frames.append(np.array(turtle.save(""))) return frames def make_gif(frames, filename): import imageio frames_rgb = [cv2.cvtColor(frame, cv2.COLOR_BGR2RGB) for frame in frames] imageio.mimsave(filename, frames_rgb, fps=30) def make_gif2(frames, filename): import imageio.v3 as iio frames_rgb = [cv2.cvtColor(frame, cv2.COLOR_BGR2RGB) for frame in frames] print(f'number of frames: {len(frames_rgb)}') iio.imwrite(filename, np.stack(frames_rgb), fps=30, plugin='pillow') def make_gif3(frames, filename): from moviepy.editor import ImageSequenceClip clip = ImageSequenceClip(list(frames), fps=20) clip.write_gif(filename, fps=20) def make_gif4(frames, filename): from array2gif import write_gif write_gif(frames, filename, fps=20) def make_gif5(frames, filename): from PIL import Image frames_rgb = [cv2.cvtColor(frame, cv2.COLOR_BGR2RGB) for frame in frames] images = [Image.fromarray(frame) for frame in frames_rgb] images[0].save(filename, save_all=True, append_images=images[1:], duration=100, loop=0) def plot4(): # the following program draws a treelike pattern import random def draw_tree(level, length, angle): if level == 0: return else: forward(length) left(angle) draw_tree(level-1, length*0.7, angle*0.8) right(angle*2) draw_tree(level-1, length*0.7, angle*0.8) left(angle) forward(-length) random.seed(0) # Comment this line to change the randomness for _ in range(7): # Adjust the number to control the density draw_tree(5, 5, 30) forward(0) left(random.randint(0, 360)) turtle.save("test3.png") return turtle.frames def plot5(): for i in range(7): with fork_state(): for j in range(4): forward(3*i) left(90.0) return turtle.frames # make_gif2(plot5(), "test.gif") frames = plot5() # frames = [cv2.cvtColor(frame, cv2.COLOR_BGR2RGB) for frame in frames] # breakpoint() # from moviepy.editor import ImageClip, concatenate_videoclips # clips = [ImageClip(frame).set_duration(1/24) for frame in frames] # concat_clip = concatenate_videoclips(clips, method="compose") # concat_clip.write_videofile("test.mp4", fps=24) img_bytes_string = turtle.save_gif("") # turtle.save('test3.png') with open("test5.gif", "wb") as f: f.write(img_bytes_string) # example_plot() # plot2()