Add examples

app.py

@@ -35,7 +35,7 @@ class Detector:
         detections = []
         result = results[0]
         for box, score, labels in zip(
-            result["boxes"], result["scores"], result["labels"]
+            result["boxes"], result["scores"], result["text_labels"]
         ):
             box = [round(x, 2) for x in box.tolist()]
             detections.append(
@@ -70,6 +70,7 @@ def detect_objects(image, labels, confidence_threshold):
 
     detections = []
     for model_name in models.keys():
+        print(f"Running detection with model: {model_name}")
         detection = models[model_name].detect(
             image,
             labels,
@@ -80,16 +81,14 @@ def detect_objects(image, labels, confidence_threshold):
     return tuple((image, det) for det in detections)
 
 
-with gr.Blocks() as demo:
+with gr.Blocks(delete_cache=(5, 10)) as demo:
     gr.Markdown("# [LLMDet](https://arxiv.org/abs/2501.18954) Arena ✨")
 
     with gr.Row():
         with gr.Column():
             gr.Markdown("## Input Image")
 
-            image_url = "http://images.cocodataset.org/val2017/000000039769.jpg"
-            image = load_image(image_url)
-            image_input = gr.Image(type="pil", image_mode="RGB", value=image)
+            image_input = gr.Image(type="pil", image_mode="RGB", format="jpeg")
 
         with gr.Column():
             gr.Markdown("## Settings")
@@ -109,19 +108,18 @@ with gr.Blocks() as demo:
                 label="Object labels (comma separated):",
                 placeholder=",".join(labels),
                 lines=1,
-                value=",".join(labels),
             )
 
     with gr.Row():
-        detect_button = gr.Button("Run Object Detection")
+        detect_button = gr.Button("Detect Objects")
 
     with gr.Row():
         gr.Markdown("## Output Annotated Images")
 
     with gr.Row():
-        output_annotated_image_tiny = gr.AnnotatedImage(label="TINY")
-        output_annotated_image_base = gr.AnnotatedImage(label="BASE")
-        output_annotated_image_large = gr.AnnotatedImage(label="LARGE")
+        output_annotated_image_tiny = gr.AnnotatedImage(label="TINY", format="jpeg")
+        output_annotated_image_base = gr.AnnotatedImage(label="BASE", format="jpeg")
+        output_annotated_image_large = gr.AnnotatedImage(label="LARGE", format="jpeg")
 
     # Connect the button to the detection function
     detect_button.click(
@@ -134,5 +132,38 @@ with gr.Blocks() as demo:
         ],
     )
 
+    with gr.Row():
+        gr.Markdown("## Examples")
+
+    with gr.Row():
+        gr.Examples(
+            examples=[
+                [
+                    "http://images.cocodataset.org/val2017/000000039769.jpg",
+                    "a cat, a remote control",
+                    0.3,
+                ],
+                [
+                    "http://images.cocodataset.org/val2017/000000370486.jpg",
+                    "a person",
+                    0.3,
+                ],
+                [
+                    "http://images.cocodataset.org/train2017/000000345263.jpg",
+                    "a red apple, a green apple",
+                    0.3,
+                ],
+            ],
+            inputs=[image_input, text_input, confidence_slider],
+            outputs=[
+                output_annotated_image_tiny,
+                output_annotated_image_base,
+                output_annotated_image_large,
+            ],
+            fn=detect_objects,
+            cache_examples=True,
+        )
+
+
 if __name__ == "__main__":
     demo.launch()