Update app.py

app.py

@@ -20,6 +20,8 @@ from depth_anything.dpt import DepthAnything
 from depth_anything.util.transform import Resize, NormalizeImage, PrepareForNet
 from moviepy.editor import *
 
+gradient = None
+params = { fnum:0, l:16, c:3, k:7 }
 frame_selected = 0
 frames = []
 depths = []
@@ -173,6 +175,7 @@ def make_video(video_path, outdir='./vis_video_depth', encoder='vits'):
         # out.release()
         cv2.destroyAllWindows()
 
+        global gradient
         global frame_selected
         global depths
         global frames
@@ -182,12 +185,15 @@ def make_video(video_path, outdir='./vis_video_depth', encoder='vits'):
         if depth_color.shape[0] == 2048: #height
             masks[len(masks)-1] = './gradient_large.png'
             depth_frames[len(masks)-1] = './gradient_large.png'
+            gradient = cv2.imread('./gradient_large.png').astype(np.uint8)
         elif depth_color.shape[0] == 1024:
             masks[len(masks)-1] = './gradient.png'
             depth_frames[len(masks)-1] = './gradient.png'
+            gradient = cv2.imread('./gradient.png').astype(np.uint8)
         else:
             masks[len(masks)-1] = './gradient_small.png'
             depth_frames[len(masks)-1] = './gradient_small.png'
+            gradient = cv2.imread('./gradient_small.png').astype(np.uint8)
         
         return final_vid, final_zip, frames, masks[frame_selected] #output_path
 
@@ -347,56 +353,60 @@ def switch_rows(v):
         print(frames[0])
         return frames
 
-def draw_mask(f, n, c, k, v, d, evt: gr.EventData):
+def draw_mask(f, l, c, k, v, d, evt: gr.EventData):
+    global params
     global frame_selected
     global masks
+    global gradient
     
     points = json.loads(v)
     pts = np.array(points, np.int32)
     pts = pts.reshape((-1,1,2))
 
-    if d["background"].shape[0] == 2048: #height
-        gradient = cv2.imread(f[0]).astype(np.uint8)
-    elif d["background"].shape[0] == 1024:
-        gradient = cv2.imread(f[1]).astype(np.uint8)
-    else:
-        gradient = cv2.imread(f[2]).astype(np.uint8)
-
-    if len(d["layers"]) == 1:
-        d["layers"].append(np.zeros_like(d["layers"][0]))
-    else:
-        d["background"] = cv2.imread(masks[frame_selected]).astype(np.uint8)
+    scale = 1
+    delta = 0
+    ddepth = cv2.CV_16S
 
-    bg = cv2.cvtColor(d["background"], cv2.COLOR_RGBA2GRAY)
+    if len(d["layers"]) == 1 or params["fnum"] != frame_selected or params["l"] != l or params["c"] != c or params["k"] != k:
+      if len(d["layers"]) == 1:
+          d["layers"].append(np.zeros_like(d["layers"][0]))
+      else:
+          d["background"] = cv2.imread(masks[frame_selected]).astype(np.uint8)
 
-    diff = (bg-cv2.cvtColor(gradient, cv2.COLOR_RGBA2GRAY)).astype(np.uint8)
-    mask = cv2.inRange(diff, -int(c/2), int(c/2))
-    #kernel = np.ones((k,k),np.float32)/(k*k)
-    #mask = cv2.filter2D(mask,-1,kernel)
+      bg = cv2.cvtColor(d["background"], cv2.COLOR_RGBA2GRAY)
 
-    indices = np.arange(0,256)   # List of all colors 
-    divider = np.linspace(0,255,n+1)[1] # we get a divider
-    quantiz = np.int0(np.linspace(0,255,n)) # we get quantization colors
-    color_levels = np.clip(np.int0(indices/divider),0,n-1) # color levels 0,1,2..
-    palette = quantiz[color_levels]
+      diff = (bg-cv2.cvtColor(gradient, cv2.COLOR_RGBA2GRAY)).astype(np.uint8)
+      mask = cv2.inRange(diff, -int(c/2), int(c/2))
+      #kernel = np.ones((k,k),np.float32)/(k*k)
+      #mask = cv2.filter2D(mask,-1,kernel)
 
-    #for i in range(n):
-    #    bg[(bg >= i*255/n) & (bg < (i+1)*255/n)] = i*255/(n-1)
-    bg = cv2.convertScaleAbs(palette[bg] + mask).astype(np.uint8) # Converting image back to uint
+      indices = np.arange(0,256)   # List of all colors 
+      divider = np.linspace(0,255,l+1)[1] # we get a divider
+      quantiz = np.int0(np.linspace(0,255,l)) # we get quantization colors
+      color_levels = np.clip(np.int0(indices/divider),0,l-1) # color levels 0,1,2..
+      palette = quantiz[color_levels]
 
-    scale = 1
-    delta = 0
-    ddepth = cv2.CV_16S
+      #for i in range(l):
+      #    bg[(bg >= i*255/l) & (bg < (i+1)*255/l)] = i*255/(l-1)
+      bg = cv2.convertScaleAbs(palette[bg] + mask).astype(np.uint8) # Converting image back to uint
 
-    grad_x = cv2.convertScaleAbs(cv2.Sobel(bg, ddepth, 1, 0, ksize=3, scale=scale, delta=delta, borderType=cv2.BORDER_DEFAULT))
-    grad_y = cv2.convertScaleAbs(cv2.Sobel(bg, ddepth, 0, 1, ksize=3, scale=scale, delta=delta, borderType=cv2.BORDER_DEFAULT))
-    grad_z = cv2.convertScaleAbs(cv2.Sobel(bg, ddepth, 1, 1, ksize=3, scale=scale, delta=delta, borderType=cv2.BORDER_DEFAULT))
+      grad_x = cv2.convertScaleAbs(cv2.Sobel(bg, ddepth, 1, 0, ksize=3, scale=scale, delta=delta, borderType=cv2.BORDER_DEFAULT))
+      grad_y = cv2.convertScaleAbs(cv2.Sobel(bg, ddepth, 0, 1, ksize=3, scale=scale, delta=delta, borderType=cv2.BORDER_DEFAULT))
+      grad_z = cv2.convertScaleAbs(cv2.Sobel(bg, ddepth, 1, 1, ksize=3, scale=scale, delta=delta, borderType=cv2.BORDER_DEFAULT))
     
-    grad = cv2.addWeighted(grad_x, 0.5, grad_y, 0.5, 0)
-    grad = cv2.addWeighted(grad, 0.5, grad_z, 0.5, 0)
+      grad = cv2.addWeighted(grad_x, 0.5, grad_y, 0.5, 0)
+      grad = cv2.addWeighted(grad, 0.5, grad_z, 0.5, 0)
+
+      mask = mask + cv2.inRange(grad, 1, 255)
+      bg[mask>0] = 255
 
-    mask = mask + cv2.inRange(grad, 1, 255)
-    bg[mask>0] = 255
+      params["fnum"] = frame_selected
+      params["l"] = l
+      params["c"] = c
+      params["k"] = k
+    else:
+      bg = cv2.cvtColor(d["layers"][0], cv2.COLOR_RGBA2GRAY)
+      mask = cv2.inRange(bg, 255, 255)
 
     x = points[len(points)-1][0]
     y = points[len(points)-1][1]