Gaussian Splatting to fail

I tried to apply Gaussian Splatting to some video frames, but the generated results looked like this, whereas the final MP4 video output had no issues. Why might this be happening? I used DA3NESTED-GIANT-LARGE-1.1.

This is my code.

```
import glob, os, torch
from depth_anything_3.api import DepthAnything3
device = torch.device("cuda:3")
model = DepthAnything3.from_pretrained("/data23/hong_yuting/Depth-Anything-3/da3_streaming/weights")
model = model.to(device=device)
example_path = "/data23/hong_yuting/Depth-Anything-3/office_images"
images = sorted(glob.glob(os.path.join(example_path, "*.png")))

output_dir = "outputs/demo"
export_format = "gs_ply"

model.inference(
images,
extrinsics=None,
intrinsics=None,
infer_gs=True,
export_dir=output_dir,
export_format=export_format,
export_feat_layers=[],
)
```