Add instance segm

app.py

@@ -47,25 +47,20 @@ def convert_to_pairs(three):
 
 
 #function to output binary mask img
-def binary_mask_img(prediction,rec_img):
+def segm_imf(prediction,rec_img):
   w= rec_img.shape[0]
   h=rec_img.shape[1]
-  # colors={"LOWER_BONE":"#fabee6","UPPER_BONE":"#96e7e6","MIDDLE_BONE":"#fffa5b"}
-  colors={"LOWER_BONE":"w","UPPER_BONE":"w","MIDDLE_BONE":"w"}
+  colors={"LOWER_BONE":"#fabee6","UPPER_BONE":"#96e7e6","MIDDLE_BONE":"#fffa5b"}
+  # colors={"LOWER_BONE":(253,226,243),"UPPER_BONE":(173,228,219),"MIDDLE_BONE": (253,247,195)}
   figure, axes = plt.subplots(figsize =(h/100.0,w/100.0))
-  img = np.zeros((w, h), dtype = np.float64)
   for prediction in prediction["predictions"]:
     points = [[p["x"], p["y"]] for p in prediction["points"]]
     polygon = patches.Polygon(
-              points, linewidth=2, edgecolor=colors[prediction["class"]],facecolor= to_rgba(colors[prediction["class"]],1),zorder=2
+              points, linewidth=2, edgecolor=colors[prediction["class"]],facecolor= to_rgba(colors[prediction["class"]],0.4)
           )
     axes.add_patch(polygon)
-  ax=plt.gca()
-  ax.invert_yaxis()
-  plt.style.use("dark_background")
-  plt.axis("off")   # turns off axes
-  plt.axis("tight")  # gets rid of white border
-  plt.axis("image")
+  plt.imshow(rec_img)
+  plt.axis("off")
   plt.tight_layout()  
   canvas = plt.get_current_fig_manager().canvas
   canvas.draw()
@@ -77,6 +72,16 @@ def binary_mask_img(prediction,rec_img):
   gray_image = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
   return gray_image
 
+def binary_cv2(rec_img,prediction):
+  w= rec_img.shape[0]
+  h=rec_img.shape[1]
+  mask = np.zeros((w, h), dtype = np.float64)
+  for prediction in prediction["predictions"]:
+    points = [[p["x"], p["y"]] for p in prediction["points"]]
+    cv2.fillPoly(mask, np.array([points]).astype(np.int32), color=(255, 0, 0))
+  masked_gray = cv2.merge((mask,mask,mask))
+  return masked_gray
+
 def extend_line(p1, p2, distance=10000):
     diff = np.arctan2(p1[1] - p2[1], p1[0] - p2[0])
     p3_x = int(p1[0] + distance*np.cos(diff))
@@ -84,20 +89,21 @@ def extend_line(p1, p2, distance=10000):
     p4_x = int(p1[0] - distance*np.cos(diff))
     p4_y = int(p1[1] - distance*np.sin(diff))
     return ((p3_x, p3_y), (p4_x, p4_y))
-def visualize(image,pred):
 
+def visualize(image,pred):
+  op_img=image.copy()
   pred=[int(i) for i in pred]
   (p1_left,p2_left)=extend_line((pred[6],pred[7]),(pred[4],pred[5]))
   (p1_h,p2_h)=extend_line((pred[4],pred[5]),(pred[2],pred[3]))
   (p1_right,p2_right)=extend_line((pred[2],pred[3]),(pred[0],pred[1]))
-  image= cv2.line(image, p1_left, (pred[4],pred[5]),(152, 216, 170),2)
-  image= cv2.line(image, p1_h, p2_h,(152, 216, 170),2)
-  image= cv2.line(image, (pred[2],pred[3]), p2_right,(152, 216, 170),2)
+  op_img= cv2.line(op_img, p1_left, (pred[4],pred[5]),(152, 216, 170),1)
+  op_img= cv2.line(op_img, p1_h, p2_h,(152, 216, 170),1)
+  op_img= cv2.line(op_img, (pred[2],pred[3]), p2_right,(152, 216, 170),1)
 
   for i in range(0,7,2):
-
-    image = cv2.circle(image, (round(pred[i]),round(pred[i+1])), int ((image.shape[0]+image.shape[1])/150), (255, 150, 128), -1)
-  return image
+    op_img = cv2.circle(op_img, (round(pred[i]),round(pred[i+1])), int ((image.shape[0]+image.shape[1])/150), (255, 150, 128), -1)
+  return op_img
+  
   
 
 
@@ -142,14 +148,15 @@ kp_predictor = DefaultPredictor(kp_cfg)
 
 def predict_fn(img_path):
     #Read and tranform input image
-    p=model.predict(img_path).json()
+    preds=model.predict(img_path).json()
     og_img=cv2.imread(img_path)
-    img=binary_mask_img(p,og_img)
+    img=binary_cv2(og_img,preds)
     outputs = kp_predictor(img)  # format is documented at https://detectron2.readthedocs.io/tutorials/models.html#model-output-format
     print("outputs==".format(outputs["instances"].to("cpu")))
     p=np.asarray(outputs["instances"].to("cpu").pred_keypoints, dtype='float32')[0].reshape(-1)
     pairss=convert_to_pairs(p)
-    output=visualize(img,pairss)
+    segm=segm_imf(preds,og_img)
+    output=visualize(segm,pairss)
     return output