| import os | |
| import sys | |
| sys.path.insert(0, os.path.abspath(os.path.join(os.path.dirname(__file__), '..', '..', '..'))) | |
| import utils3d | |
| import numpy as np | |
| import torch | |
| def run(): | |
| for i in range(100): | |
| if i == 0: | |
| spatial = [] | |
| else: | |
| dim = np.random.randint(4) | |
| spatial = [np.random.randint(1, 10) for _ in range(dim)] | |
| fovy = np.random.uniform(5 / 180 * np.pi, 175 / 180 * np.pi, spatial) | |
| aspect = np.random.uniform(0.01, 100, spatial) | |
| near = np.random.uniform(0.1, 100, spatial) | |
| far = np.random.uniform(near*2, 1000, spatial) | |
| expected = utils3d.numpy.perspective(fovy, aspect, near, far) | |
| device = [torch.device('cpu'), torch.device('cuda')][np.random.randint(2)] | |
| fovy = torch.tensor(fovy, device=device) | |
| aspect = torch.tensor(aspect, device=device) | |
| near = torch.tensor(near, device=device) | |
| far = torch.tensor(far, device=device) | |
| actual = utils3d.torch.perspective(fovy, aspect, near, far).cpu().numpy() | |
| assert np.allclose(expected, actual), '\n' + \ | |
| 'Input:\n' + \ | |
| f'\tfovy: {fovy}\n' + \ | |
| f'\taspect: {aspect}\n' + \ | |
| f'\tnear: {near}\n' + \ | |
| f'\tfar: {far}\n' + \ | |
| 'Actual:\n' + \ | |
| f'{actual}\n' + \ | |
| 'Expected:\n' + \ | |
| f'{expected}' | |