add test for kornia.geometry.ransac

.github/workflows/update-s2s.yml

@@ -16,9 +16,9 @@ jobs:
                  test_metrics, test_morphology, test_nerf, test_sensors, test_utils, geometry/test_bbox,
                  geometry/test_boxes, geometry/test_calibration, geometry/test_camera, geometry/test_conversions,
                  geometry/test_depth, geometry/test_epipolar, geometry/test_homography, geometry/test_liegroup,
-                 geometry/test_linalg, geometry/test_line, geometry/test_quaternion, geometry/test_solvers, geometry/test_subpix,
-                 geometry/test_transform, geometry/test_vector, augmentation/test_augmentation, augmentation/test_auto,
-                 augmentation/test_container ]
+                 geometry/test_linalg, geometry/test_line, geometry/test_quaternion, geometry/test_ransac,
+                 geometry/test_solvers, geometry/test_subpix, geometry/test_transform, geometry/test_vector,
+                 augmentation/test_augmentation, augmentation/test_auto, augmentation/test_container ]
         compile : [ F ]
     steps:
       - name: Checkout ivy-integration-tests

kornia/geometry/test_ransac.py

@@ -0,0 +1,83 @@
+from helpers import (
+    _check_allclose,
+    _check_shape_allclose,
+    _nest_array_to_numpy,
+    _nest_torch_tensor_to_new_framework,
+)
+
+import ivy
+import kornia
+import pytest
+import torch
+
+
+# Helpers #
+# ------- #
+
+def _to_numpy_and_shape_allclose(torch_x, transpiled_x, tolerance=1e-3):
+    orig_data = _nest_array_to_numpy(torch_x)
+    transpiled_data = _nest_array_to_numpy(transpiled_x)
+    _check_shape_allclose(orig_data, transpiled_data, tolerance=tolerance) 
+
+
+# Tests #
+# ----- #
+
+def test_RANSAC(target_framework, mode, backend_compile):
+    print("kornia.geometry.ransac.RANSAC")
+
+    if backend_compile:
+        pytest.skip()
+
+    # Initialize RANSAC with default parameters
+    ransac = kornia.geometry.ransac.RANSAC(model_type='homography')
+
+    # Transpile the RANSAC class to the target framework
+    TranspiledRANSAC = ivy.transpile(kornia.geometry.ransac.RANSAC, source="torch", target=target_framework)
+
+    # Prepare synthetic keypoints data for source and destination images
+    kp1 = torch.tensor([[0.0, 0.0], [1.0, 1.0], [2.0, 2.0], [3.0, 3.0]], requires_grad=True)
+    kp2 = torch.tensor([[0.0, 0.0], [1.1, 1.1], [2.0, 2.1], [3.0, 3.1]], requires_grad=True)
+
+    transpiled_kp1 = _nest_torch_tensor_to_new_framework(kp1, target_framework)
+    transpiled_kp2 = _nest_torch_tensor_to_new_framework(kp2, target_framework)
+
+    # Run RANSAC on the original data
+    torch_model, torch_inliers = ransac(kp1, kp2)
+
+    # Run transpiled RANSAC on the transpiled data
+    transpiled_ransac = TranspiledRANSAC(model_type='homography')
+    transpiled_model, transpiled_inliers = transpiled_ransac(transpiled_kp1, transpiled_kp2)
+
+    # Ensure that the estimated models are close to each other
+    _to_numpy_and_shape_allclose(torch_model, transpiled_model)
+
+    # Ensure that the inlier masks are consistent
+    _to_numpy_and_shape_allclose(torch_inliers, transpiled_inliers)
+
+    # Test RANSAC with custom parameters
+    ransac_custom = kornia.geometry.ransac.RANSAC(model_type='homography', inl_th=1.5, max_iter=20, confidence=0.95)
+    TranspiledRANSACCustom = ivy.transpile(kornia.geometry.ransac.RANSAC, source="torch", target=target_framework)
+    transpiled_ransac_custom = TranspiledRANSACCustom(model_type='homography', inl_th=1.5, max_iter=20, confidence=0.95)
+
+    torch_model_custom, torch_inliers_custom = ransac_custom(kp1, kp2)
+    transpiled_model_custom, transpiled_inliers_custom = transpiled_ransac_custom(transpiled_kp1, transpiled_kp2)
+
+    _to_numpy_and_shape_allclose(torch_model_custom, transpiled_model_custom)
+    _to_numpy_and_shape_allclose(torch_inliers_custom, transpiled_inliers_custom)
+
+    # Test RANSAC on a different model type (e.g., fundamental matrix)
+    ransac_fundamental = kornia.geometry.ransac.RANSAC(model_type='fundamental')
+    TranspiledRANSACFundamental = ivy.transpile(kornia.geometry.ransac.RANSAC, source="torch", target=target_framework)
+    transpiled_ransac_fundamental = TranspiledRANSACFundamental(model_type='fundamental')
+
+    kp1 = torch.tensor([[0.0, 0.0], [1.0, 1.0], [2.0, 2.0], [3.0, 3.0], [0.0, 0.0], [1.0, 1.0], [2.0, 2.0], [3.0, 3.0]], requires_grad=True)
+    kp2 = torch.tensor([[0.0, 0.0], [1.1, 1.1], [2.0, 2.1], [3.0, 3.1], [0.0, 0.0], [1.1, 1.1], [2.0, 2.1], [3.0, 3.1]], requires_grad=True)
+    transpiled_kp1 = _nest_torch_tensor_to_new_framework(kp1, target_framework)
+    transpiled_kp2 = _nest_torch_tensor_to_new_framework(kp2, target_framework)
+
+    torch_model_fundamental, torch_inliers_fundamental = ransac_fundamental(kp1, kp2)
+    transpiled_model_fundamental, transpiled_inliers_fundamental = transpiled_ransac_fundamental(transpiled_kp1, transpiled_kp2)
+
+    _to_numpy_and_shape_allclose(torch_model_fundamental, transpiled_model_fundamental)
+    _to_numpy_and_shape_allclose(torch_inliers_fundamental, transpiled_inliers_fundamental)