Merge pull request #82 from JuliaImages/vs/features2d-wrappers

Wrap features2d (2/2): generated wrappers + tests (closes #51)

Author: ViralBShah

src/cv_manual_wrap.jl

@@ -28,12 +28,6 @@ function CascadeClassifier(filename::String)
 end
 
 
-function detect(cobj::cv_Ptr{T}, image::InputArray, mask::InputArray) where {T <: Feature2D}
-	return cpp_to_julia(jlopencv_cv_cv_Feature2D_cv_Feature2D_detect(julia_to_cpp(cobj),julia_to_cpp(image),julia_to_cpp(mask)))
-end
-detect(cobj::cv_Ptr{T}, image::InputArray; mask::InputArray = (CxxMat())) where {T <: Feature2D} = detect(cobj, image, mask)
-
-
 function detectMultiScale(cobj::CascadeClassifier, image::InputArray, scaleFactor::Float64, minNeighbors::Int32, flags::Int32, minSize::Size{Int32}, maxSize::Size{Int32})
 	return cpp_to_julia(jlopencv_cv_cv_CascadeClassifier_cv_CascadeClassifier_detectMultiScale(julia_to_cpp(cobj),julia_to_cpp(image),julia_to_cpp(scaleFactor),julia_to_cpp(minNeighbors),julia_to_cpp(flags),julia_to_cpp(minSize),julia_to_cpp(maxSize)))
 end
@@ -43,11 +37,6 @@ function empty(cobj::CascadeClassifier)
 	return cpp_to_julia(jlopencv_cv_cv_CascadeClassifier_cv_CascadeClassifier_empty(julia_to_cpp(cobj)))
 end
 
-function SimpleBlobDetector_create(parameters::SimpleBlobDetector_Params)
-	return cpp_to_julia(jlopencv_cv_cv_SimpleBlobDetector_create(julia_to_cpp(parameters)))
-end
-SimpleBlobDetector_create(; parameters::SimpleBlobDetector_Params = (SimpleBlobDetector_Params())) = SimpleBlobDetector_create(parameters)
-
 ## Convenience: mirror Python's `cv.VideoWriter_fourcc(*"h264")`. Issue #31.
 function VideoWriter_fourcc(s::AbstractString)
     length(s) == 4 ||

test/test_feature2d.jl

@@ -1,26 +1,23 @@
-# test simple blob detector
-# TODO: the keypoint distance assertion below is intermittently wrong across
-# platforms, and on Julia 1.11 + Windows `OpenCV.detect` itself hits a
-# stack-overflow in the artifact's `julia_to_cpp` stride fallback. Disable the
-# whole block until both issues are addressed.
-#
-# img_gray = OpenCV.imread(joinpath(test_dir, "shared", "pic1.png"), OpenCV.IMREAD_GRAYSCALE)
-#
-# detector = OpenCV.SimpleBlobDetector_create()
-#
-# # Compare centers of keypoints and se how many of them match,
-# kps = OpenCV.detect(detector, img_gray)
-#
-# kps_expect = [OpenCV.Point{Float32}(174.9114f0, 227.75146f0),OpenCV.Point{Float32}(106.925545f0, 179.5765f0)]
-# for kp in kps
-#     closest_match = 100000
-#     for kpe in kps_expect
-#         dx = kpe.x - kp.pt.x
-#         dy = kpe.y - kp.pt.y
-#         if sqrt(dx*dx+dy*dy) < closest_match
-#             closest_match = sqrt(dx*dx+dy*dy)
-#         end
-#     end
-#
-#     @test closest_match < 10
-# end
+@testset "features2d" begin
+    img_gray = OpenCV.imread(joinpath(test_dir, "shared", "pic1.png"), OpenCV.IMREAD_GRAYSCALE)
+
+    @testset "ORB detect+compute+match" begin
+        orb = OpenCV.ORB_create()
+        kps = OpenCV.detect(orb, img_gray)
+        @test length(kps) > 0
+        kps2, desc = OpenCV.compute(orb, img_gray, kps)
+        @test length(kps2) == length(kps)
+        @test size(desc, 1) == length(kps)
+        @test size(desc, 2) == OpenCV.descriptorSize(orb)
+
+        bf = OpenCV.BFMatcher_create(OpenCV.NORM_HAMMING, true)
+        matches = OpenCV.match(bf, desc, desc)
+        @test length(matches) == length(kps)
+    end
+
+    @testset "SimpleBlobDetector" begin
+        detector = OpenCV.SimpleBlobDetector_create()
+        kps = OpenCV.detect(detector, img_gray)
+        @test length(kps) > 0
+    end
+end