Update app.py

app.py

@@ -39,7 +39,6 @@ blurin = "1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
 edge = []
 gradient = None
 params = { "fnum":0 }
-dcolor = []
 pcolors = []
 frame_selected = 0
 frames = []
@@ -459,7 +458,6 @@ def seg_frame(newmask, b, d):
 
 
 def select_frame(d, evt: gr.SelectData):
-    global dcolor
     global frame_selected
     global depths
     global masks
@@ -468,13 +466,8 @@ def select_frame(d, evt: gr.SelectData):
     if evt.index != frame_selected:
         edge = []
         frame_selected = evt.index
-
-    if len(dcolor) == 0:
-        bg = [127, 127, 127, 255]
-    else:
-        bg = "[" + str(dcolor[frame_selected])[1:-1] + ", 255]"
         
-    return depths[frame_selected], frame_selected, bg
+    return depths[frame_selected], frame_selected
 
 def switch_rows(v):
     global frames
@@ -486,61 +479,6 @@ def switch_rows(v):
         print(frames[0])
         return frames
 
-def optimize(v, d):
-    global pcolors
-    global dcolor
-    global frame_selected
-    global frames
-    global depths
-    
-    if v == True:
-        ddepth = cv2.CV_16S
-        kernel_size = 3
-        l = 16
-
-        dcolor = []
-        for k, f in enumerate(frames):
-            frame = cv2.imread(frames[k]).astype(np.uint8)
-            
-            # convert to np.float32
-            f = np.float32(frame.reshape((-1,3)))
-            # define criteria, number of clusters(K) and apply kmeans()
-            criteria = (cv2.TERM_CRITERIA_EPS + cv2.TERM_CRITERIA_MAX_ITER, 4, 1.0)
-            ret,label,center=cv2.kmeans(f,l,None,criteria,4,cv2.KMEANS_RANDOM_CENTERS)
-            # Now convert back into uint8, and make original image
-            center = np.uint8(center)
-            res = center[label.flatten()]
-            frame = res.reshape((frame.shape))
-
-            depth = cv2.imread(depths[k]).astype(np.uint8)
-            mask = cv2.cvtColor(depth, cv2.COLOR_RGB2GRAY)
-            dcolor.append(bincount(frame[mask==0]))
-            print(dcolor[k])
-            clrs = Image.fromarray(frame.astype(np.uint8)).convert('RGB').getcolors()
-            i=0
-            while i<len(clrs):
-                clrs[i] = list(clrs[i][1])
-                clrs[i].append(255)
-                i=i+1
-            print(clrs)
-            pcolors = clrs
-            
-            #mask = cv2.convertScaleAbs(cv2.Laplacian(cv2.cvtColor(frame, cv2.COLOR_RGB2GRAY), ddepth, ksize=kernel_size))
-            #mask[mask>0] = 255
-            #frame[mask==0] = (0, 0, 0)
-            cv2.imwrite(frames[k], frame)
-
-            #depth[mask==0] = (255,255,255)
-            mask = cv2.inRange(frame, np.array([dcolor[k][0]-8, dcolor[k][1]-8, dcolor[k][2]-8]), np.array([dcolor[k][0]+8, dcolor[k][1]+8, dcolor[k][2]+8]))
-            depth[mask>0] = (255,255,255)
-            depth[depth.shape[0]-1:depth.shape[0], 0:depth.shape[1]] = (160, 160, 160)
-            depth[0:1, 0:depth.shape[1]] = (0, 0, 0)
-            cv2.imwrite(depths[k], depth)
-            
-    if d == False:
-      return frames, "[" + str(dcolor[frame_selected])[1:-1] + ", 255]"
-    else:
-      return depths, "[" + str(dcolor[frame_selected])[1:-1] + ", 255]"
 
 def bincount(a):
     a2D = a.reshape(-1,a.shape[-1])
@@ -560,7 +498,7 @@ def reset_mask(d):
     backup = cv2.imread(backups[frame_selected]).astype(np.uint8)
     cv2.imwrite(frames[frame_selected], backup)
 
-    d["layers"][0] = np.zeros(d["layers"][0].shape[:2],np.uint8)
+    d["layers"][0] = None
 
     return gr.ImageEditor(value=d)
 
@@ -645,7 +583,7 @@ def findNormals(format):
 
 
 load_model="""
-async(c, o, b, p, d, n, m)=>{
+async(c, o, p, d, n, m)=>{
   var intv = setInterval(function(){
     if (document.getElementById("iframe3D")===null || typeof document.getElementById("iframe3D")==="undefined") {
       try {
@@ -666,12 +604,8 @@ async(c, o, b, p, d, n, m)=>{
             }
           });
           
-          var bg = JSON.parse(document.getElementById("bgcolor").getElementsByTagName("textarea")[0].value);
           BABYLON.Engine.LastCreatedScene.getEngine().setHardwareScalingLevel(1.0);
-          for (var i=0; i<bg.length; i++) {
-            bg[i] /= 255;
-          }
-          BABYLON.Engine.LastCreatedScene.clearColor = new BABYLON.Color4(bg[0], bg[1], bg[2], bg[3]);
+          BABYLON.Engine.LastCreatedScene.clearColor = new BABYLON.Color4(255,255,255,255);
           BABYLON.Engine.LastCreatedScene.ambientColor = new BABYLON.Color4(255,255,255,255);
           //BABYLON.Engine.LastCreatedScene.autoClear = false;
           //BABYLON.Engine.LastCreatedScene.autoClearDepthAndStencil = false;
@@ -748,7 +682,6 @@ async(c, o, b, p, d, n, m)=>{
         var o_ = o.split(",");
         document.getElementById("iframe3D").contentDocument.getElementById("coords").value = c;
         document.getElementById("iframe3D").contentDocument.getElementById("order").value = o;
-        document.getElementById("iframe3D").contentDocument.getElementById("bgcolor").value = b;
         document.getElementById("iframe3D").contentDocument.getElementById("bgimage").value = "";
         document.getElementById("iframe3D").contentDocument.getElementById("bgdepth").value = "";
         for (var j=0; j<o_.length; j++) {
@@ -918,9 +851,6 @@ with gr.Blocks(css=css, js=js) as demo:
             with gr.Accordion(label="Depths", open=False):
               output_depth = gr.Files(label="Depth files", interactive=False)
             output_switch.input(fn=switch_rows, inputs=[output_switch], outputs=[output_frame])
-            optimize_switch = gr.Checkbox(label="Optimize")
-            bgcolor = gr.Textbox(elem_id="bgcolor", value="[127, 127, 127, 255]", label="Background color", interactive=False)
-            optimize_switch.input(fn=optimize, inputs=[optimize_switch, output_switch], outputs=[output_frame, bgcolor])
             output_mask = gr.ImageEditor(layers=False, sources=('upload', 'clipboard'), show_download_button=True, type="numpy", interactive=True, transforms=(None,), eraser=gr.Eraser(), brush=gr.Brush(default_size=0, colors=['black', '#505050', '#a0a0a0', 'white']), elem_id="image_edit")
             with gr.Row():
               selector = gr.HTML(value="""
@@ -1034,7 +964,7 @@ with gr.Blocks(css=css, js=js) as demo:
                 blur_in = gr.Textbox(elem_id="blur_in", label="Kernel size", show_label=False, interactive=False, value=blurin)
             with gr.Accordion(label="Locations", open=False):
                 selected = gr.Number(elem_id="fnum", value=0, minimum=0, maximum=256, interactive=False)
-                output_frame.select(fn=select_frame, inputs=[output_mask], outputs=[output_mask, selected, bgcolor])
+                output_frame.select(fn=select_frame, inputs=[output_mask], outputs=[output_mask, selected])
                 example_coords = """[
                   {"lat": 50.07379596793083, "lng": 14.437146122950555, "heading": 152.70303, "pitch": 2.607833999999997}, 
                   {"lat": 50.073799567020004, "lng": 14.437146774240507, "heading": 151.12973, "pitch": 2.8672300000000064}, 
@@ -1174,7 +1104,7 @@ with gr.Blocks(css=css, js=js) as demo:
         return output_video_path + (json.dumps(locations),)
 
     submit.click(on_submit, inputs=[input_video, model_type, blur_in, boffset, bsize, coords], outputs=[processed_video, processed_zip, output_frame, output_mask, output_depth, coords])
-    render.click(None, inputs=[coords, mesh_order, bgcolor, output_frame, output_mask, selected, output_depth], outputs=None, js=load_model)
+    render.click(None, inputs=[coords, mesh_order, output_frame, output_mask, selected, output_depth], outputs=None, js=load_model)
     render.click(partial(get_mesh), inputs=[output_frame, output_mask, blur_in, load_all], outputs=[result, result_file, mesh_order])
 
     example_files = [["./examples/streetview.mp4", "vits", blurin, 1, 32, example_coords]]