Mirrored video results for DyNeRF dataset

[ktkt2020]

Hi,

Thanks for your nice work!

Here is a small issue about the mirrored video results (flipped the left and right). It will appear when we rendering the DyNeRF dataset.

I found that the poses returned by average_poses and viewmatrix in 4DGaussians/scene/neural_3D_dataset_NDC.py were left-handed.
Should we change these to right-handed instead?
i.e.,

    # 4. Compute the x axis
    x = normalize(np.cross(y_, z))  # (3)

    # 5. Compute the y axis (as z and x are normalized, y is already of norm 1)
    y = np.cross(z, x)  # (3)

and

    m[:3] = np.stack([vec0, vec1, vec2, pos], 1)

[guanjunwu]

Thank you for your suggestion!
Rendering validation video will not affect experiment results in my paper and only for visualization. And I'll fix it in the next version.