Cannot run out of the box google collab pixelNERF

[1ssb]

There seems to a data.zip file for pixel Nerf implementation. Do you mind listing out the steps to get it and more importantly a list of steps to perform inference on a new image, directly from the collab?

[airalcorn2]

The instructions for generating the ShapeNet datasets can be found in the README. The final cell of the Colab notebook demonstrates how to perform inference on a test image and viewpoint:

test_W_i = E(test_source_image)
(_, C_rs_f) = run_one_iter_of_pixelnerf(
    test_ds,
    N_c,
    t_i_c_bin_edges,
    t_i_c_gap,
    test_os,
    camera_distance,
    scale,
    test_W_i,
    chunk_size,
    F_c,
    N_f,
    t_f,
    N_d,
    d_std,
    t_n,
    F_f,
)
plt.imshow(C_rs_f.detach().cpu().numpy())

[1ssb]

Hi Michael, the very first block of code is not working, the data.zip has not been provided directly, is the shapenet data, the data.zip?

On Thu, 10 Aug, 2023, 9:41 pm Michael A. Alcorn, @.***> wrote:

Closed #8 https://github.com/airalcorn2/pytorch-nerf/issues/8 as completed.

— Reply to this email directly, view it on GitHub https://github.com/airalcorn2/pytorch-nerf/issues/8#event-10058369669, or unsubscribe https://github.com/notifications/unsubscribe-auth/AJWHFEDR73EHSFMD7QC4W3LXUTCHZANCNFSM6AAAAAA3K5WQBI . You are receiving this because you authored the thread.Message ID: @.***>

[airalcorn2]

You need to generate the data yourself using the instructions provided in the README, zip the generated data folder, and then put that in your own Google Drive.

[1ssb]

Thanks, how would you like to be cited?

[airalcorn2]

No citation necessary. You can just link to this repository in any projects that build off of it.

[1ssb]

Hi is there a way to train the pixelNerf using this implementation?

[airalcorn2]

The script/notebook trains a pixelNeRF.

[1ssb]

Awesome, thanks!

On Mon, 14 Aug, 2023, 10:00 pm Michael A. Alcorn, @.***> wrote:

The script/notebook trains a pixelNeRF.

— Reply to this email directly, view it on GitHub https://github.com/airalcorn2/pytorch-nerf/issues/8#issuecomment-1677190939, or unsubscribe https://github.com/notifications/unsubscribe-auth/AJWHFEEFHSQHZIM6HNT55JTXVIHN3ANCNFSM6AAAAAA3K5WQBI . You are receiving this because you authored the thread.Message ID: @.***>

[Reznox7]

@airalcorn2 Can we train custom datasets using any of the approaches you provided?