update app

app.py

@@ -268,22 +268,17 @@ with gr.Blocks() as demo:
     
     <h3 style="color: #2a9d8f;">Get Started</h3>
     To get started quickly, you can click on our <strong>examples (at the bottom of the page)</strong>. The example results are cached, allowing you to view them even when in a queue.
-
     If you want to reconstruct your own data, simply **(a)** upload images (.jpg, .png, etc.) or **(b)** upload a video (.mp4, .mov, etc.).
-
     <div style="font-size: 16px; line-height: 1.2;">
     <h3 style="color: #2a9d8f;">Hyperparameters</h3>
     Typically, 4 query images and 2048 query points are sufficient. For a denser point cloud, use 4096 query points. If the reconstruction appears incomplete, increase to 6 query images. Note that excessive queries can lead to out-of-memory errors.
     
     <h3 style="color: #2a9d8f;">Converting Video to Frames</h3>
     By default, we convert the input video to frames at 1 frame per second. To prevent hugging face space crashes, we limit reconstruction to the first 25 frames. If both images and videos are uploaded, the demo will only reconstruct the uploaded images.
-
     <h3 style="color: #2a9d8f;">Dynamic Scenes</h3>
     SfM methods are designed for rigid/static reconstruction. When dealing with dynamic/moving inputs, these methods may still work by focusing on the rigid parts of the scene. However, to ensure high-quality results, it is better to minimize the presence of moving objects in the input data.
-
     <h3 style="color: #2a9d8f;">Runtime</h3>
     The reconstruction typically takes <strong>up to 90 seconds</strong>. Longer runtimes can be attributed to difficult input data or a high number of query images/points. Please note that running reconstruction on Hugging Face is slower than on a local machine.
-
     <h3 style="color: #2a9d8f;">Contact</h3>
     If you meet any problem, feel free to create an issue in our <a href="https://github.com/facebookresearch/vggsfm" target="_blank" style="color: #2a9d8f;">GitHub Repo</a> ⭐
     </div>