fix: A lot of fixes relating to bounding-boxes

src/facedet/retinaface.rs

@@ -62,9 +62,18 @@ pub struct FaceDetectionModelOutput {
     pub landmark: ndarray::Array3<f32>,
 }
 
+/// Represents the 5 facial landmarks detected by RetinaFace
+pub struct FaceLandmarks {
+    pub left_eye: Point2<f32>,
+    pub right_eye: Point2<f32>,
+    pub nose: Point2<f32>,
+    pub left_mouth: Point2<f32>,
+    pub right_mouth: Point2<f32>,
+}
 pub struct FaceDetectionProcessedOutput {
     pub bbox: Vec<Aabb2<f32>>,
     pub confidence: Vec<f32>,
+    pub landmarks: Vec<FaceLandmarks>,
 }
 
 impl FaceDetectionModelOutput {
@@ -87,10 +96,12 @@ impl FaceDetectionModelOutput {
         }
         let mut boxes = Vec::new();
         let mut scores = Vec::new();
+        let mut landmarks = Vec::new();
         let var0 = config.variance[0];
         let var1 = config.variance[1];
         let bbox_data = self.bbox;
         let conf_data = self.confidence;
+        let landmark_data = self.landmark;
         let num_priors = bbox_data.shape()[1];
         for idx in 0..num_priors {
             let dx = bbox_data[[0, idx, 0]];
@@ -110,11 +121,35 @@ impl FaceDetectionModelOutput {
             if score > config.threshold {
                 boxes.push(Aabb2::from_x1y1x2y2(x_min, y_min, x_max, y_max));
                 scores.push(score);
+
+                let left_eye_x = landmark_data[[0, idx, 0]] * anchor_w * var0 + anchor_cx;
+                let left_eye_y = landmark_data[[0, idx, 1]] * anchor_h * var0 + anchor_cy;
+
+                let right_eye_x = landmark_data[[0, idx, 2]] * anchor_w * var0 + anchor_cx;
+                let right_eye_y = landmark_data[[0, idx, 3]] * anchor_h * var0 + anchor_cy;
+
+                let nose_x = landmark_data[[0, idx, 4]] * anchor_w * var0 + anchor_cx;
+                let nose_y = landmark_data[[0, idx, 5]] * anchor_h * var0 + anchor_cy;
+
+                let left_mouth_x = landmark_data[[0, idx, 6]] * anchor_w * var0 + anchor_cx;
+                let left_mouth_y = landmark_data[[0, idx, 7]] * anchor_h * var0 + anchor_cy;
+
+                let right_mouth_x = landmark_data[[0, idx, 8]] * anchor_w * var0 + anchor_cx;
+                let right_mouth_y = landmark_data[[0, idx, 9]] * anchor_h * var0 + anchor_cy;
+
+                landmarks.push(FaceLandmarks {
+                    left_eye: Point2::new(left_eye_x, left_eye_y),
+                    right_eye: Point2::new(right_eye_x, right_eye_y),
+                    nose: Point2::new(nose_x, nose_y),
+                    left_mouth: Point2::new(left_mouth_x, left_mouth_y),
+                    right_mouth: Point2::new(right_mouth_x, right_mouth_y),
+                });
             }
         }
         Ok(FaceDetectionProcessedOutput {
             bbox: boxes,
             confidence: scores,
+            landmarks,
         })
     }
 }