Add python api that generates desciptor(s) from the aligned image(s) (#1667)

* Add python api that generates desciptor(s) from the aligned image(s)

* Remove asserts from face_recognition.py example/tutorial

* In batch_compute_face_descriptors_from_aligned_images, use for-in loop to simplify the code

Improvde the document on binding methods and the error message if the aligned image is not of size 150x150

python_examples/face_recognition.py

@@ -114,6 +114,23 @@ for f in glob.glob(os.path.join(faces_folder_path, "*.jpg")):
         # In particular, a padding of 0.5 would double the width of the cropped area, a value of 1.
         # would triple it, and so forth.
 
+        # There is another overload of compute_face_descriptor that can take
+        # as an input an aligned image. 
+        #
+        # Note that it is important to generate the aligned image as
+        # dlib.get_face_chip would do it i.e. the size must be 150x150, 
+        # centered and scaled.
+        #
+        # Here is a sample usage of that
+
+        print("Computing descriptor on aligned image ..")
+        
+        # Let's generate the aligned image using get_face_chip
+        face_chip = dlib.get_face_chip(img, shape)        
+
+        # Now we simply pass this chip (aligned image) to the api
+        face_descriptor_from_prealigned_image = facerec.compute_face_descriptor(face_chip)                
+        print(face_descriptor_from_prealigned_image)        
         
         dlib.hit_enter_to_continue()
 

tools/python/src/face_recognition.cpp

@@ -43,6 +43,15 @@ public:
         return compute_face_descriptors(img, faces, num_jitters, padding)[0];
     }
 
+    matrix<double,0,1> compute_face_descriptor_from_aligned_image (
+        numpy_image<rgb_pixel> img,        
+        const int num_jitters
+    )
+    {
+        std::vector<numpy_image<rgb_pixel>> images{img};        
+        return batch_compute_face_descriptors_from_aligned_images(images, num_jitters)[0];                
+    }
+
     std::vector<matrix<double,0,1>> compute_face_descriptors (
         numpy_image<rgb_pixel> img,
         const std::vector<full_object_detection>& faces,
@@ -127,6 +136,52 @@ public:
         return face_descriptors;
     }
 
+    std::vector<matrix<double,0,1>> batch_compute_face_descriptors_from_aligned_images (
+        const std::vector<numpy_image<rgb_pixel>>& batch_imgs,        
+        const int num_jitters
+    )
+    {
+        dlib::array<matrix<rgb_pixel>> face_chips;           
+        for (auto& img : batch_imgs) {
+
+            matrix<rgb_pixel> image;
+            if (is_image<unsigned char>(img))
+                assign_image(image, numpy_image<unsigned char>(img));
+            else if (is_image<rgb_pixel>(img))
+                assign_image(image, numpy_image<rgb_pixel>(img));
+            else
+                throw dlib::error("Unsupported image type, must be 8bit gray or RGB image.");
+
+            // Check for the size of the image
+            if ((image.nr() != 150) || (image.nc() != 150)) {
+                throw dlib::error("Unsupported image size, it should be of size 150x150. Also cropping must be done as `dlib.get_face_chip` would do it. \
+                That is, centered and scaled essentially the same way.");
+            }
+
+            face_chips.push_back(image);        
+        }       
+
+        std::vector<matrix<double,0,1>> face_descriptors;
+        if (num_jitters <= 1)
+        {
+            // extract descriptors and convert from float vectors to double vectors
+            auto descriptors = net(face_chips, 16);      
+
+            for (auto& des: descriptors) {
+                face_descriptors.push_back(matrix_cast<double>(des));
+            }       
+        }
+        else
+        {
+            // extract descriptors and convert from float vectors to double vectors
+            for (auto& fimg : face_chips) {
+                auto& r = mean(mat(net(jitter_image(fimg, num_jitters), 16)));
+                face_descriptors.push_back(matrix_cast<double>(r)); 
+            }
+        }        
+        return face_descriptors;        
+    }
+
 private:
 
     dlib::rand rnd;
@@ -300,6 +355,12 @@ void bind_face_recognition(py::module &m)
             "If num_jitters>1 then each face will be randomly jittered slightly num_jitters times, each run through the 128D projection, and the average used as the face descriptor. "
             "Optionally allows to override default padding of 0.25 around the face."
             )
+        .def("compute_face_descriptor", &face_recognition_model_v1::compute_face_descriptor_from_aligned_image,
+            py::arg("img"), py::arg("num_jitters")=0,
+            "Takes an aligned face image of size 150x150 and converts it into a 128D face descriptor."
+            "Note that the alignment should be done in the same way dlib.get_face_chip does it."
+            "If num_jitters>1 then image will be randomly jittered slightly num_jitters times, each run through the 128D projection, and the average used as the face descriptor. "            
+            )
         .def("compute_face_descriptor", &face_recognition_model_v1::compute_face_descriptors,
             py::arg("img"), py::arg("faces"), py::arg("num_jitters")=0, py::arg("padding")=0.25,
             "Takes an image and an array of full_object_detections that reference faces in that image and converts them into 128D face descriptors.  "
@@ -309,9 +370,16 @@ void bind_face_recognition(py::module &m)
         .def("compute_face_descriptor", &face_recognition_model_v1::batch_compute_face_descriptors,
             py::arg("batch_img"), py::arg("batch_faces"), py::arg("num_jitters")=0, py::arg("padding")=0.25,
             "Takes an array of images and an array of arrays of full_object_detections. `batch_faces[i]` must be an array of full_object_detections corresponding to the image `batch_img[i]`, "
-            "referencing faces in that image. Every face will be converting into 128D face descriptors.  "
+            "referencing faces in that image. Every face will be converted into 128D face descriptors.  "
             "If num_jitters>1 then each face will be randomly jittered slightly num_jitters times, each run through the 128D projection, and the average used as the face descriptor. "
             "Optionally allows to override default padding of 0.25 around the face."
+            )
+        .def("compute_face_descriptor", &face_recognition_model_v1::batch_compute_face_descriptors_from_aligned_images,
+            py::arg("batch_img"), py::arg("num_jitters")=0,
+            "Takes an array of aligned images of faces of size 150_x_150."
+            "Note that the alignment should be done in the same way dlib.get_face_chip does it."
+            "Every face will be converted into 128D face descriptors.  "
+            "If num_jitters>1 then each face will be randomly jittered slightly num_jitters times, each run through the 128D projection, and the average used as the face descriptor. "            
             );
     }