specified tokens

app.py

@@ -63,8 +63,12 @@ glip_demo = GLIPDemo(
 )
 
 
-def predict(image, text):
-    result, _ = glip_demo.run_on_web_image(image[:, :, [2, 1, 0]], text, 0.5)
+def predict(image, text, specified_tokens=""):
+    if specified_tokens.strip() == "":
+        specified_tokens = None
+    else:
+        specified_tokens = specified_tokens..split(";")
+    result, _ = glip_demo.run_on_web_image(image[:, :, [2, 1, 0]], text, 0.5, specified_tokens)
     return result[:, :, [2, 1, 0]]
 
 
@@ -73,7 +77,7 @@ image = gr.inputs.Image()
 gr.Interface(
     description="Object Recognition with DesCo",
     fn=predict,
-    inputs=["image", "text"],
+    inputs=["image", "text", "ground tokens (optional)"],
     outputs=[
         gr.outputs.Image(
             type="pil",
@@ -83,9 +87,10 @@ gr.Interface(
     examples=[
         #["./flickr_9472793441.jpg", "bobble heads on top of the shelf ."],
         #["./flickr_9472793441.jpg", "sofa . remote . dog . person . car . sky . plane ."],
-        ["./coco_000000281759.jpg", "A green umbrella. A pink striped umbrella. A plain white umbrella."],
-        ["./coco_000000281759.jpg", "a flowery top. A blue dress. An orange shirt ."],
-        ["./coco_000000281759.jpg", "a car . An electricity box ."],
+        ["./coco_000000281759.jpg", "A green umbrella. A pink striped umbrella. A plain white umbrella.", ""],
+        ["./coco_000000281759.jpg", "a flowery top. A blue dress. An orange shirt .", ""],
+        ["./coco_000000281759.jpg", "a car . An electricity box .", ""],
+        ["./1.jpg", "a train besides sidewalk", "train;sidewalk"],
         #["./flickr_7520721.jpg", "A woman figure skater in a blue costume holds her leg by the blade of her skate ."]
     ],
     article=Path("docs/intro.md").read_text()