Enhance predict function to return targets not found and update main.py to display results in the UI

inference_utils/init_predict_mock.py

@@ -11,9 +11,49 @@
 # limitations under the License.
 
 
+from typing import Tuple
+from PIL import Image, ImageDraw, ImageFont
+import gradio as gr
+import random
+
+
 def init_model():
     return None
 
 
-def predict(image, modality_type: str, targets: list[str]):
-    return image
+def predict(
+    image: Image, modality_type: str, targets: list[str]
+) -> Tuple[gr.Image, str]:
+    # Randomly split targets into found and not found
+    targets_found = random.sample(targets, k=len(targets) // 2)
+    targets_not_found = [t for t in targets if t not in targets_found]
+
+    # Create a copy of the image to draw on
+    image_with_text = image.copy()
+    draw = ImageDraw.Draw(image_with_text)
+
+    # Draw found targets on the image with larger font
+    font_size = 36
+    try:
+        font = ImageFont.truetype("DejaVuSans.ttf", font_size)
+    except OSError:
+        # Fallback to default font if DejaVuSans is not available
+        font = ImageFont.load_default()
+
+    # Calculate starting position from bottom
+    line_height = 50
+    total_height = len(targets_found) * line_height
+    padding = 20
+
+    # Start from bottom and work upwards
+    y_position = image_with_text.height - total_height - padding
+    for target in targets_found:
+        draw.text((20, y_position), target, fill="red", font=font)
+        y_position += line_height
+
+    # Format targets_not_found as a string with one target per line
+    targets_not_found_str = (
+        "\n".join(targets_not_found) if targets_not_found else "All targets were found!"
+    )
+
+    return image_with_text, targets_not_found_str

main.py

@@ -114,6 +114,9 @@ def run():
                 )
             with gr.Column():
                 output_image = gr.Image(type="pil", label="Prediction")
+                output_targets_not_found = gr.Textbox(
+                    label="Targets Not Found", lines=4, max_lines=10
+                )
 
         input_modality_type.change(
             fn=update_input_targets,
@@ -125,7 +128,7 @@ def run():
         submit_btn.click(
             fn=predict,
             inputs=[input_image, input_modality_type, input_targets],
-            outputs=output_image,
+            outputs=[output_image, output_targets_not_found],
         )
 
         gr.Examples(