init

app.py

@@ -142,16 +142,22 @@ def camera_understanding(image_src, question, seed, progress=gr.Progress(track_t
 
     img_up = Image.open(saved_paths[0]).convert("RGB")
     img_lat = Image.open(saved_paths[1]).convert("RGB")
-    
     w, h = img_up.size
     left = max(0, w - h)
     img_up = img_up.crop((left, 0, w, h))
-
     w, h = img_lat.size
     left = max(0, w - h)
     img_lat = img_lat.crop((left, 0, w, h))
+    img_up = img_up.resize((512, 512))
+    img_lat = img_lat.resize((512, 512))
+    
+    gap = 10
+    W, H = img_up.size
+    combined = Image.new("RGB", (W * 2 + gap, H), (255, 255, 255))
+    combined.paste(img_up, (0, 0))
+    combined.paste(img_lat, (W + gap, 0))
 
-    return text, [img_up], [img_lat]
+    return text, combined
 
 
 @torch.inference_mode()
@@ -249,8 +255,7 @@ with gr.Blocks(css=css) as demo:
         understanding_button = gr.Button("Chat")
         understanding_output = gr.Textbox(label="Response")
         
-        camera1 = gr.Gallery(label="Camera Maps (Up Vector)", columns=1, rows=1)
-        camera2 = gr.Gallery(label="Camera Maps (Latitude)", columns=1, rows=1)
+        camera_map = gr.Image(label="Camera Maps")
 
         with gr.Accordion("Advanced options", open=False):
             und_seed_input = gr.Number(label="Seed", precision=0, value=42)
@@ -277,7 +282,7 @@ with gr.Blocks(css=css) as demo:
     understanding_button.click(
         camera_understanding,
         inputs=[image_input, und_seed_input],
-        outputs=[understanding_output, camera1, camera2]
+        outputs=[understanding_output, camera_map]
     )
 
 demo.launch(share=True)