Download identifying_shapes/script.py from aggr8/Percept-V: direct link, hf CLI and curl.
- Browser
- Download file 3.85 kB
-
https://huggingface.co/datasets/aggr8/Percept-V/resolve/main/identifying_shapes/script.py
- Command line
-
hf download hf://datasets/aggr8/Percept-V/identifying_shapes/script.py
-
curl -L -o script.py https://huggingface.co/datasets/aggr8/Percept-V/resolve/main/identifying_shapes/script.py
3.85 kB
| import matplotlib.pyplot as plt | |
| import matplotlib.patches as mp | |
| import random | |
| import os | |
| import json | |
| import sys | |
| import argparse | |
| def create_images(num_images, num_objects_list, items_list): | |
| # Ensure the output directory exists | |
| output_dir = os.path.join(os.getcwd(), "data") | |
| if not os.path.exists(output_dir): | |
| os.makedirs(output_dir) | |
| data = [] | |
| file = 1 | |
| for num_objects in num_objects_list: | |
| dpi = 100 # Dots per inch | |
| width_inches = 1 # 100 pixels | |
| height_inches = num_objects # 100 pixels per row | |
| for img_index in range(num_images): | |
| fig, ax = plt.subplots(figsize=(width_inches, height_inches), dpi=dpi) | |
| image_data = [] | |
| image_side = [] | |
| for i in range (num_objects): | |
| shape, sides = random.choice(list(items_list.items())) | |
| y = num_objects - i - 1 | |
| image_data.append(shape) | |
| image_side.append(sides) | |
| # Ensure the circles do not overlap by adjusting positions | |
| # Add circles to the plot | |
| if(shape == 'Circle'): | |
| circle = mp.Circle((0.5, y), 0.3, color='blue', fill=True) | |
| ax.add_patch(circle) | |
| if(shape == 'Diamond'): | |
| sq = mp.RegularPolygon((0.5, y), numVertices=4, radius=0.3, facecolor='green') | |
| ax.add_patch(sq) | |
| if(shape == 'Triangle'): | |
| tri = mp.RegularPolygon((0.5, y), numVertices=3, radius=0.3, facecolor='red') | |
| ax.add_patch(tri) | |
| if(shape == 'Pentagon'): | |
| pent = mp.RegularPolygon((0.5, y), numVertices=5, radius=0.3, facecolor='cyan') | |
| ax.add_patch(pent) | |
| if(shape == 'Hexagon'): | |
| hex = mp.RegularPolygon((0.5, y), numVertices=6, radius=0.3, facecolor='magenta') | |
| ax.add_patch(hex) | |
| if i < num_objects -1: | |
| ax.plot([0, 1], [y - 0.5, y - 0.5], color='black') | |
| ax.set_xlim(0, 1) | |
| ax.set_ylim(-0.5, num_objects - 0.5) | |
| ax.set_aspect('equal') | |
| ax.axis('off') | |
| image_filename = f"{file}.png" | |
| file += 1 | |
| image_path = os.path.join(output_dir, image_filename) | |
| plt.savefig(image_path, bbox_inches='tight', dpi=dpi) | |
| plt.close(fig) | |
| # Store the number of objects in the image_data dictionary | |
| gold_output = { | |
| "id": image_filename, | |
| "Gold_output": image_data, | |
| "Gold_side": image_side, | |
| "Rows" : num_objects | |
| } | |
| data.append(gold_output) | |
| # Save the data to a JSON file | |
| with open(os.path.join(os.getcwd() , "data.json"), "w") as json_file: | |
| json.dump(data, json_file, indent=4) | |
| if __name__ == "__main__": | |
| # Parse command line arguments | |
| parser = argparse.ArgumentParser(description='Create a grid of circles and triangles.') | |
| parser.add_argument( | |
| '--num_images', | |
| type=int, | |
| help='Number of images to generate', | |
| default=1 | |
| ) | |
| parser.add_argument( | |
| '--num_sizes', | |
| nargs='*', | |
| type=int, | |
| help='List of count of circles', | |
| default=[5] | |
| ) | |
| parser.add_argument( | |
| '--file', | |
| type=int, | |
| help='Starting file number', | |
| default=1 | |
| ) | |
| args = parser.parse_args() | |
| file = args.file | |
| num_object_list = args.num_sizes | |
| append = (file != 1) | |
| items_list = {'Circle':0, 'Triangle':3, 'Diamond':4, 'Pentagon':5, 'Hexagon':6} | |
| create_images(args.num_images, num_object_list, items_list) | |
| # print(num_object_list) | |
| # Create images with the specified number of objects |