Update app.py

app.py

@@ -131,7 +131,7 @@ def recognize_image(image, texts):
     output_map = (output_map - output_map.min()) / (output_map.max() - output_map.min())
     heatmap = show_cam_on_image(img_d, output_map, use_rgb=True)
 
-    return Image.fromarray(heatmap), {texts.split(';')[i]: float(prediction[i]) for i in range(len(texts.split(';')))}
+    return {texts.split(';')[i]: float(prediction[i]) for i in range(len(texts.split(';')))}, Image.fromarray(img_d), Image.fromarray(heatmap)
 
 
 image = gr.inputs.Image()
@@ -142,10 +142,13 @@ gr.Interface(
     fn=recognize_image,
     inputs=["image", "text"],
     outputs=[                          
+        label, 
+        gr.outputs.Image(
+        type="pil",
+        label="crop input"),           
         gr.outputs.Image(
         type="pil",
         label="zero-shot heat map"),           
-        label
     ],
     examples=[
     ["./elephants.png", "an elephant; an elephant walking in the river; four elephants walking in the river"],