init

app.py

@@ -111,8 +111,8 @@ def camera_understanding(image_src, question, seed, progress=gr.Progress(track_t
 @spaces.GPU(duration=120)  # Specify a duration to avoid timeout
 def generate_image(prompt_scene,
                    seed=42,
-                   roll=3,
-                   pitch=1.0,
+                   roll=0.1,
+                   pitch=0.1,
                    fov=1.0,
                    progress=gr.Progress(track_tqdm=True)):
     # Clear CUDA cache and avoid tracking gradients
@@ -134,6 +134,7 @@ def generate_image(prompt_scene,
     cam_map = cam_map / (math.pi / 2)
     
     prompt = prompt_scene + " " + prompt_camera
+    print("prompt:", prompt)
     
     bsz = 4
     with torch.no_grad():
@@ -175,7 +176,7 @@ with gr.Blocks(css=css) as demo:
                 roll = gr.Slider(minimum=-0.7854, maximum=0.7854, value=0.1000, step=0.1000, label="roll value")
                 pitch = gr.Slider(minimum=-0.7854, maximum=0.7854, value=-0.1000, step=0.1000, label="pitch value")
                 fov = gr.Slider(minimum=0.3491, maximum=1.8326, value=1.5000, step=0.1000, label="fov value")
-            seed_input = gr.Number(label="Seed (Optional)", precision=0, value=1234)
+            seed_input = gr.Number(label="Seed (Optional)", precision=0, value=42)
             
         generation_button = gr.Button("Generate Images")