import importlib.util import json import sys from pathlib import Path import pytest import torch def _load_panorama_module(): path = Path(__file__).parents[1] / "utils" / "panorama.py" spec = importlib.util.spec_from_file_location("panorama_utils_under_test", path) module = importlib.util.module_from_spec(spec) sys.modules[spec.name] = module spec.loader.exec_module(module) return module _PANORAMA = _load_panorama_module() equirectangular_to_perspective = _PANORAMA.equirectangular_to_perspective normalize_panorama_view = _PANORAMA.normalize_panorama_view def coordinate_panorama(width: int, height: int) -> torch.Tensor: u = (torch.arange(width, dtype=torch.float32) + 0.5) / width v = (torch.arange(height, dtype=torch.float32) + 0.5) / height grid_v, grid_u = torch.meshgrid(v, u, indexing="ij") zeros = torch.zeros_like(grid_u) return torch.stack((grid_u, grid_v, zeros), dim=-1).unsqueeze(0) def panorama_view( yaw: float = 0, pitch: float = 0, hfov: float = 90, aspect_ratio: float = 1, ) -> dict: return { "version": 1, "projection": "equirectangular", "yaw": yaw, "pitch": pitch, "hfov": hfov, "aspect_ratio": aspect_ratio, } def center_pixel(image: torch.Tensor) -> torch.Tensor: return image[0, image.shape[1] // 2, image.shape[2] // 2] def test_normalize_panorama_view_matches_frontend_ranges(): assert normalize_panorama_view(panorama_view(yaw=540, pitch=100, hfov=10)) == { "version": 1, "projection": "equirectangular", "yaw": -180.0, "pitch": 100.0, "hfov": 30.0, "aspect_ratio": 1.0, } assert normalize_panorama_view(panorama_view(pitch=270))["pitch"] == -90.0 def test_default_view_samples_panorama_center(): output = equirectangular_to_perspective( coordinate_panorama(width=720, height=360), panorama_view(), width=101, height=101, ) assert output.shape == (1, 101, 101, 3) assert center_pixel(output)[0].item() == pytest.approx(0.5, abs=0.01) assert center_pixel(output)[1].item() == pytest.approx(0.5, abs=0.01) @pytest.mark.parametrize(("yaw", "expected_u"), [(90, 0.75), (-90, 0.25)]) def test_yaw_moves_center_horizontally(yaw: float, expected_u: float): output = equirectangular_to_perspective( coordinate_panorama(720, 360), panorama_view(yaw=yaw), width=101, height=101, ) assert center_pixel(output)[0].item() == pytest.approx(expected_u, abs=0.01) @pytest.mark.parametrize(("pitch", "expected_v"), [(45, 0.25), (-45, 0.75)]) def test_pitch_moves_center_vertically(pitch: float, expected_v: float): output = equirectangular_to_perspective( coordinate_panorama(720, 360), panorama_view(pitch=pitch), width=101, height=101, ) assert center_pixel(output)[1].item() == pytest.approx(expected_v, abs=0.01) def test_sampling_wraps_across_horizontal_seam(): output = equirectangular_to_perspective( coordinate_panorama(720, 360), panorama_view(yaw=179, hfov=30), width=101, height=101, ) center_row_u = output[0, 50, :, 0] assert torch.isfinite(center_row_u).all() assert center_row_u.min().item() < 0.1 assert center_row_u.max().item() > 0.9 def test_same_aspect_ratio_preserves_key_rays_across_resolutions(): panorama = coordinate_panorama(720, 360) small = equirectangular_to_perspective( panorama, panorama_view(aspect_ratio=2), width=101, height=51, ) large = equirectangular_to_perspective( panorama, panorama_view(aspect_ratio=2), width=201, height=101, ) assert torch.allclose(center_pixel(small), center_pixel(large), atol=0.01, rtol=0) assert torch.allclose(small[0, 0, 0], large[0, 0, 0], atol=0.02, rtol=0) def test_golden_camera_centers_match_typescript_contract(): cases = json.loads( (Path(__file__).parent / "fixtures" / "panorama_camera_cases.json").read_text(encoding="utf-8") ) panorama = coordinate_panorama(720, 360) for case in cases: output = equirectangular_to_perspective( panorama, panorama_view(yaw=case["yaw"], pitch=case["pitch"]), width=101, height=101, ) center = center_pixel(output) assert center[0].item() == pytest.approx(case["u"], abs=0.01), case["name"] assert center[1].item() == pytest.approx(case["v"], abs=0.01), case["name"] @pytest.mark.parametrize( "image, width, height", [ (torch.zeros(10, 20, 3), 10, 10), (torch.zeros(1, 10, 20, 3), 0, 10), (torch.zeros(1, 10, 20, 3), 10, -1), ], ) def test_projection_rejects_invalid_shapes_and_sizes(image, width, height): with pytest.raises(ValueError): equirectangular_to_perspective(image, panorama_view(), width, height)