Added test.py, updated requirements.txt

requirements.txt

@@ -1,13 +1,13 @@
 thop==0.0.31.post2005241907
 joblib==1.0.1
-torchvision==0.10.0+cu111
+torchvision>=0.10.0+cu111
 tqdm==4.50.2
-torch==1.9.0+cu111
+torch>=1.9.0+cu111
 matplotlib==3.3.2
 requests==2.25.1
 numpy==1.20.3
 opencv_python==4.5.2.52
-scipy==1.5.2
+scipy>=1.5.2
 pandas==1.1.3
 seaborn==0.11.1
 onnx==1.9.0

test.py

@@ -0,0 +1,43 @@
+from face_detector import YoloDetector
+import numpy as np
+from PIL import Image
+import cv2
+
+def show_results(img, xywh, landmarks):
+    h,w,c = img.shape
+    tl = 5 or round(0.2 * (h + w) / 2) + 1  # line/font thickness
+
+    n = 0
+    for i in xywh:
+        x1 =i[0]
+        y1 = i[1]
+        x2 = i[2]
+        y2 = i[3]
+        cv2.rectangle(img, (x1,y1), (x2, y2), (0,255,0), thickness=tl, lineType=cv2.LINE_AA)
+
+        clors = [(255,0,0),(0,255,0),(0,0,255),(255,255,0),(0,255,255)]
+
+        c=0
+        for j in landmarks[n]:
+            cv2.circle(img, (j[0],j[1]), tl+1, clors[c], -1)
+            c=c+1
+
+        tf = max(tl - 1, 1)  # font thickness    
+        cv2.putText(img, 'Face {0}'.format(n), (x1, y1 - 2), 0, tl / 3, [225, 255, 255], thickness=tf, lineType=cv2.LINE_AA)
+        n=n+1
+
+    return img
+# gpu=-1 means use cpu, gpu=n where n is gpu device id
+model = YoloDetector(target_size=720,gpu=-1,min_face=90)
+img = cv2.imread('test_image.jpg')
+orgimg = np.array(img)
+bboxes,points = model.predict(orgimg)
+
+faces = show_results(img,bboxes,points)
+win_name = "visualization"
+#Named window for fit-to-window
+cv2.namedWindow(win_name, cv2.WINDOW_NORMAL)
+cv2.imshow(win_name,faces)
+cv2.waitKey(0)
+cv2.destroyAllWindows()
+