Update app.py

app.py

@@ -130,8 +130,9 @@ def recognize_image(image, texts):
     output_map = output_map.squeeze(1).detach().permute(1, 2, 0).numpy()
     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 {texts.split(';')[i]: float(prediction[i]) for i in range(len(texts.split(';')))}, Image.fromarray(np.uint8(255 * img_d)), Image.fromarray(heatmap)
+    
+    show_img = np.concatenate((np.uint8(255 * img_d), heatmap), 1)
+    return {texts.split(';')[i]: float(prediction[i]) for i in range(len(texts.split(';')))}, Image.fromarray(show_img)
 
 
 image = gr.inputs.Image()
@@ -145,10 +146,7 @@ gr.Interface(
         label, 
         gr.outputs.Image(
         type="pil",
-        label="crop input"),           
-        gr.outputs.Image(
-        type="pil",
-        label="zero-shot heat map"),           
+        label="crop input/heat map"),                
     ],
     examples=[
     ["./elephants.png", "an elephant; an elephant walking in the river; four elephants walking in the river"],