import numpy as np import itertools as it import os from PIL import Image from random import random from tex_utils import * from mobject import * class ImageMobject(Mobject2D): """ Automatically filters out black pixels """ # SHOULD_BUFF_POINTS = False def __init__(self, image_file, filter_color = "black", invert = True, use_cache = True, *args, **kwargs): Mobject2D.__init__(self, *args, **kwargs) self.filter_rgb = 255 * np.array(Color(filter_color).get_rgb()).astype('uint8') self.name = to_cammel_case( os.path.split(image_file)[-1].split(".")[0] ) self.use_cache = use_cache possible_paths = [ image_file, os.path.join(IMAGE_DIR, image_file), os.path.join(IMAGE_DIR, image_file + ".jpg"), os.path.join(IMAGE_DIR, image_file + ".png"), ] for path in possible_paths: if os.path.exists(path): self.generate_points_from_file(path, invert) return raise IOError("File not Found") def generate_points_from_file(self, path, invert): if self.use_cache and self.read_in_cached_attrs(path, invert): return image = Image.open(path).convert('RGB') if invert: image = invert_image(image) self.generate_points_from_image_array(np.array(image)) self.cache_attrs(path, invert) def get_cached_attr_files(self, path, invert, attrs): #Hash should be unique to (path, invert) pair unique_hash = str(hash(path+str(invert))) return [ os.path.join(IMAGE_MOBJECT_DIR, unique_hash)+"."+attr for attr in attrs ] def read_in_cached_attrs(self, path, invert, attrs = ("points", "rgbs"), dtype = "float64"): cached_attr_files = self.get_cached_attr_files(path, invert, attrs) if all(map(os.path.exists, cached_attr_files)): for attr, cache_file in zip(attrs, cached_attr_files): arr = np.fromfile(cache_file, dtype = dtype) arr = arr.reshape(arr.size/self.DIM, self.DIM) setattr(self, attr, arr) return True return False def cache_attrs(self, path, invert, attrs = ("points", "rgbs"), dtype = "float64"): cached_attr_files = self.get_cached_attr_files(path, invert, attrs) for attr, cache_file in zip(attrs, cached_attr_files): getattr(self, attr).astype(dtype).tofile(cache_file) def generate_points_from_image_array(self, image_array): height, width = image_array.shape[:2] #Flatten array, and find indices where rgb is not filter_rgb array = image_array.reshape((height * width, 3)) bools = array == self.filter_rgb bools = bools[:,0]*bools[:,1]*bools[:,2] indices = np.arange(height * width, dtype = 'int')[~bools] rgbs = array[indices, :].astype('float') / 255.0 points = np.zeros((indices.size, 3), dtype = 'float64') points[:,0] = indices%width - width/2 points[:,1] = -indices/width + height/2 height, width = map(float, (height, width)) if height / width > float(DEFAULT_HEIGHT) / DEFAULT_WIDTH: points *= 2 * SPACE_HEIGHT / height else: points *= 2 * SPACE_WIDTH / width self.add_points(points, rgbs = rgbs) def should_buffer_points(self): # potentially changed in subclasses return False class Face(ImageMobject): def __init__(self, mode = "simple", *args, **kwargs): """ Mode can be "simple", "talking", "straight" """ ImageMobject.__init__(self, mode + "_face", *args, **kwargs) self.scale(0.5) self.center() class VideoIcon(ImageMobject): def __init__(self, *args, **kwargs): ImageMobject.__init__(self, "video_icon", *args, **kwargs) self.scale(0.3) self.center() def text_mobject(text, size = None): size = size or "\\Large" #TODO, auto-adjust? return tex_mobject(text, size, TEMPLATE_TEXT_FILE) def tex_mobject(expression, size = None, template_tex_file = TEMPLATE_TEX_FILE): if size == None: if len("".join(expression)) < MAX_LEN_FOR_HUGE_TEX_FONT: size = "\\Huge" else: size = "\\large" #Todo, make this more sophisticated. image_files = tex_to_image(expression, size, template_tex_file) if isinstance(image_files, list): #TODO, is checking listiness really the best here? result = CompoundMobject(*map(ImageMobject, image_files)) else: result = ImageMobject(image_files) return result.highlight("white").center()