| import numpy as np |
| import json |
| import pdb |
| from matplotlib import pyplot as plt |
| import os |
|
|
| |
| from scipy.ndimage import binary_dilation, binary_erosion, binary_hit_or_miss |
| import random |
|
|
| from ListSelEm import * |
| from Utils import Process, Change_Colour |
|
|
|
|
| def generate_color_change_rule(no_colors): |
| """ |
| """ |
| arr = np.zeros((2**no_colors, no_colors+1), dtype=np.int32) |
| for i in range(2**no_colors): |
| str_binary = ("0"*no_colors + bin(i)[2:])[-no_colors:] |
| arr[i, :-1] = np.array([int(x) for x in str_binary]) |
| arr[:, -1] = np.random.randint(1, no_colors+1, 2**no_colors) |
| arr[0, -1] = 0 |
| return arr |
|
|
|
|
| def generate_inp_out_catA_Hard(list_se_idx, color_rule, **param): |
| """ |
| """ |
| base_img = np.zeros((param['img_size'], param['img_size']), dtype=np.int32) |
| sz = np.random.randint(2, 4) |
| for color in range(1, param['no_colors']+1): |
| idx1 = np.random.randint(0, param['img_size'], size=sz) |
| idx2 = np.random.randint(0, param['img_size'], size=sz) |
| base_img[idx1, idx2] = color |
|
|
| |
| base_img = Process(base_img, num_colors=param['no_colors']) |
| for color in range(param['no_colors']): |
| idx = np.random.randint(0, 8) |
| base_img[:, :, color] = binary_dilation(base_img[:, :, color], list_se_3x3[idx]) |
| base_img = Change_Colour(base_img, rule=None) |
|
|
| inp_img = np.array(base_img, copy=True) |
| out_img = np.array(base_img, copy=True) |
| out_img = Process(out_img, num_colors=param['no_colors']) |
|
|
| for (color, list_se_color) in zip(range(0, param['no_colors']), list_se_idx): |
| for idx in list_se_color: |
| out_img[:, :, color] = binary_dilation(out_img[:, :, color], list_se_3x3[idx]) |
|
|
| for (color, list_se_color) in zip(range(0, param['no_colors']), list_se_idx): |
| for idx in list_se_color: |
| out_img[:, :, color] = binary_erosion(out_img[:, :, color], list_se_3x3[idx]) |
|
|
| out_img = Change_Colour(out_img, color_rule) |
|
|
| return inp_img, out_img |
|
|
|
|
| def generate_one_task_CatA_Hard(**param): |
| """ |
| """ |
| list_se_idx = [] |
| for _ in range(param['no_colors']): |
| list_se_idx.append(np.random.randint(0, 8, 4)) |
|
|
| color_rule = generate_color_change_rule(param['no_colors']) |
|
|
| data = [] |
| k = 0 |
| while k < param['no_examples_per_task']: |
| inp_img, out_img = generate_inp_out_catA_Hard(list_se_idx, color_rule, **param) |
|
|
| |
| FLAG = False |
|
|
| for col in range(param['no_colors']+1): |
| if np.all(inp_img == col): |
| FLAG = True |
| if np.all(out_img == col): |
| FLAG = True |
|
|
| if FLAG: |
| |
| list_se_idx = [] |
| for _ in range(param['no_colors']): |
| list_se_idx.append(np.random.randint(0, 8, 4)) |
|
|
| color_rule = generate_color_change_rule(param['no_colors']) |
| data = [] |
| k = -1 |
| else: |
| |
| data.append((inp_img, out_img)) |
| k += 1 |
|
|
| return data, list_se_idx, color_rule |
|
|
|
|
| def write_dict_json_CatA_Hard(data, fname): |
| """ |
| """ |
| dict_data = [] |
| for (inp, out) in data: |
| inp = [[int(y) for y in x] for x in inp] |
| out = [[int(y) for y in x] for x in out] |
| dict_data.append({"input": inp, "output": out}) |
|
|
| with open(fname, "w") as f: |
| f.write(json.dumps(dict_data)) |
|
|
|
|
| def write_solution_CatA_Hard(list_se_idx, color_rule, fname): |
| """ |
| """ |
| with open(fname, 'w') as f: |
| band = 1 |
| for list_se_color in list_se_idx: |
| f.write("Sequence for Band {}\n".format(band)) |
| f.write("---------------------- \n") |
| for idx in list_se_color: |
| f.write("Dilation SE{}\n".format(idx+1)) |
| for idx in list_se_color: |
| f.write("Erosion SE{}\n".format(idx+1)) |
|
|
| band += 1 |
|
|
| f.write("\n") |
|
|
| f.write("\n Color Change Rule \n") |
| f.write("------------------\n") |
| f.write(json.dumps([[int(y) for y in x] for x in color_rule])) |
|
|
|
|
| def write_solution_CatA_Hard_json(list_se_idx, color_rule, fname): |
| """ |
| Solution written in format: |
| band - 1/2/3/None |
| op - Dilation/Erosion/Color_Change |
| SE = SE0-SE7 |
| """ |
| data = [] |
| band = 1 |
| for list_se_color in list_se_idx: |
| for idx in list_se_color: |
| data.append((band, 'Dilation', 'SE{}'.format(idx+1))) |
| for idx in list_se_color: |
| data.append((band, 'Erosion', 'SE{}'.format(idx+1))) |
| band += 1 |
|
|
| data.append((None, 'color_rule', (([[int(y) for y in x] for x in color_rule])))) |
| with open(fname, "w") as f: |
| f.write(json.dumps(data)) |
|
|
|
|
| def generate_100_tasks_CatA_Hard(seed, **param): |
| """ |
| """ |
| np.random.seed(seed) |
| os.makedirs("./Dataset/CatA_Hard", exist_ok=True) |
| for task_no in range(100): |
| data, list_se_idx, color_rule = generate_one_task_CatA_Hard(**param) |
| fname = './Dataset/CatA_Hard/Task{:03d}.json'.format(task_no) |
| write_dict_json_CatA_Hard(data, fname) |
|
|
| fname = './Dataset/CatA_Hard/Task{:03d}_soln.txt'.format(task_no) |
| write_solution_CatA_Hard(list_se_idx, color_rule, fname) |
|
|
| fname = './Dataset/CatA_Hard/Task{:03d}_soln.json'.format(task_no) |
| write_solution_CatA_Hard_json(list_se_idx, color_rule, fname) |
|
|
|
|
| if __name__ == "__main__": |
| param = {} |
| param['img_size'] = 15 |
| param['se_size'] = 5 |
| param['seq_length'] = 4 |
| param['no_examples_per_task'] = 4 |
| param['no_colors'] = 3 |
|
|
| generate_100_tasks_CatA_Hard(32, **param) |
|
|