Add crop_box to inference ops

abraia/inference/detect.py

@@ -4,7 +4,7 @@
 import numpy as np
 import onnxruntime as ort
 
-from .ops import py_cpu_nms, normalize, mask_to_polygon, softmax, count_objects
+from .ops import non_maximum_suppression, normalize, mask_to_polygon, softmax, count_objects
 from ..utils import download_file, load_json, get_color, get_providers
 from ..utils import load_image, show_image, save_image, Video, render_results
 
@@ -15,14 +15,6 @@ def resize(img, size):
     return cv2.resize(img, (width, height), interpolation=cv2.INTER_LINEAR)
 
 
-# TODO: Refactor to convert to get detection
-def crop(img, size):
-    height, width = img.shape[:2]
-    left, top = (width - size) // 2, (height - size) // 2
-    right, bottom = left + size, top + size
-    return img[top:bottom, left:right]
-
-
 def preprocess(img, size = 224):
     img = resize(img, size)
     img = normalize(img, [0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
@@ -52,36 +44,6 @@ def prepare_input(img, shape):
     return img / 255
 
 
-def iou(box1, box2):
-    """Calculates the intersection-over-union of two boxes."""
-    tl1, wh1, br1 = [box1[0], box1[1]], [box1[2], box1[3]], [box1[0] + box1[2], box1[1] + box1[3]]
-    tl2, wh2, br2 = [box2[0], box2[1]], [box2[2], box2[3]], [box2[0] + box2[2], box2[1] + box2[3]]
-    intersection_area = np.prod(np.maximum(np.minimum(br1, br2) - np.maximum(tl1, tl2), 0))
-    union_area = np.prod(wh1) + np.prod(wh2) - intersection_area;
-    return intersection_area / union_area
-
-
-# def non_maximum_suppression(objects, iou_threshold):
-#     results = []
-#     objects.sort(key=lambda obj: obj['confidence'], reverse=True)
-#     while len(objects) > 0:
-#         results.append(objects[0])
-#         objects = [obj for obj in objects if iou(obj['box'], objects[0]['box']) < iou_threshold]
-    # return results
-
-
-def non_maximum_suppression(objects, iou_threshold):
-    dets = []
-    for obj in objects:
-        s = obj['confidence']
-        x, y, w, h = obj['box']
-        dets.append([x, y, x + w, y + h, s])
-    if dets:
-        idxs = py_cpu_nms(np.array(dets), iou_threshold)
-        return [objects[idx] for idx in idxs]
-    return []
-
-
 def sigmoid_mask(z):
     mask = 1 / (1 + np.exp(-z))
     return (255 * mask).astype('uint8')

abraia/inference/ops.py

@@ -126,6 +126,24 @@ def py_cpu_nms(dets, thresh):
     return keep
 
 
+# def iou(box1, box2):
+#     """Calculates the intersection-over-union of two boxes."""
+#     tl1, wh1, br1 = [box1[0], box1[1]], [box1[2], box1[3]], [box1[0] + box1[2], box1[1] + box1[3]]
+#     tl2, wh2, br2 = [box2[0], box2[1]], [box2[2], box2[3]], [box2[0] + box2[2], box2[1] + box2[3]]
+#     intersection_area = np.prod(np.maximum(np.minimum(br1, br2) - np.maximum(tl1, tl2), 0))
+#     union_area = np.prod(wh1) + np.prod(wh2) - intersection_area;
+#     return intersection_area / union_area
+
+
+# def non_maximum_suppression(objects, iou_threshold):
+#     results = []
+#     objects.sort(key=lambda obj: obj['confidence'], reverse=True)
+#     while len(objects) > 0:
+#         results.append(objects[0])
+#         objects = [obj for obj in objects if iou(obj['box'], objects[0]['box']) < iou_threshold]
+    # return results
+
+
 def non_maximum_suppression(objects, iou_threshold):
     dets = []
     for obj in objects:
@@ -152,3 +170,8 @@ def count_objects(results):
         colors[label] = color
     objects = [{'label': label, 'count': counts[label], 'color': colors[label]} for label in counts.keys()]
     return objects
+
+
+def crop_box(img, box):
+    x, y, w, h = box
+    return img[y:y+w, x:x+w]

abraia/inference/plates.py

@@ -15,6 +15,7 @@ def iou(box1, box2):
     return intersection_area / union_area
 
 
+# TODO: Replace with non_maximum_suppression from ops
 def nms(objects, iou_threshold = 0.5):
     results = []
     objects.sort(key=lambda obj: obj['confidence'], reverse=True)

abraia/tracker/matching.py

@@ -110,19 +110,6 @@ def embedding_distance(tracks, detections, metric='cosine'):
     return cost_matrix
 
 
-# def gate_cost_matrix(kf, cost_matrix, tracks, detections, only_position=False):
-#     if cost_matrix.size == 0:
-#         return cost_matrix
-#     gating_dim = 2 if only_position else 4
-#     gating_threshold = kalman_filter.chi2inv95[gating_dim]
-#     measurements = np.asarray([det.to_xyah() for det in detections])
-#     for row, track in enumerate(tracks):
-#         gating_distance = kf.gating_distance(
-#             track.mean, track.covariance, measurements, only_position)
-#         cost_matrix[row, gating_distance > gating_threshold] = np.inf
-#     return cost_matrix
-
-
 def fuse_motion(kf, cost_matrix, tracks, detections, only_position=False, lambda_=0.98):
     if cost_matrix.size == 0:
         return cost_matrix