Skip to main content

Implementation of D4RT, Efficiently Reconstructing Dynamic Scenes

Project description

d4rt (wip)

Implementation of D4RT, Efficiently Reconstructing Dynamic Scenes, Deepmind

install

$ pip install d4rt

usage

from torch import randn, randint
from d4rt.d4rt import D4RT

model = D4RT(
    dim = 512,
    video_image_size = 128,
    video_patch_size = 32,
    video_max_time_len = 10,
    enc_depth = 6,
    dec_depth = 6
)

videos = randn(2, 10, 3, 128, 128)

video_lens = randint(2, 10, (2,)) # handle variable lengthed video, can be None for max length always

# inputs

coors = randint(0, 128, (2, 5, 2))
time_src = randint(0, 10, (2, 5))
time_tgt = randint(0, 10, (2, 5))
time_camera = randint(0, 10, (2, 5))

query_lens = randint(1, 5, (2,)) # handle varaible lengthed queries

# output

points = randn(2, 5, 3)

loss = model(
    videos,
    video_lens = video_lens,
    coors = coors,
    time_src = time_src,
    time_tgt = time_tgt,
    time_camera = time_camera,
    query_lens = query_lens,
    points = points,
)

loss.backward()

# without giving the output, it returns the prediction

pred = model(
    videos,
    coors = coors,
    time_src = time_src,
    time_tgt = time_tgt,
    time_camera = time_camera
)

assert pred.shape == (2, 5, 3)

citations

@article{zhang2025d4rt,
    title   = {Efficiently Reconstructing Dynamic Scenes One D4RT at a Time},
    author  = {Zhang, Chuhan and Le Moing, Guillaume and Koppula, Skanda and Rocco, Ignacio and Momeni, Liliane and Xie, Junyu and Sun, Shuyang and Sukthankar, Rahul and Barral, Jo{\"e}lle K. and Hadsell, Raia and Ghahramani, Zoubin and Zisserman, Andrew and Zhang, Junlin and Sajjadi, Mehdi S. M.},
    journal = {arXiv preprint},
    year    = {2025}
}
@inproceedings{liu2026geometryaware,
    title   = {Geometry-aware 4D Video Generation for Robot Manipulation},
    author  = {Zeyi Liu and Shuang Li and Eric Cousineau and Siyuan Feng and Benjamin Burchfiel and Shuran Song},
    booktitle = {The Fourteenth International Conference on Learning Representations},
    year    = {2026},
    url     = {https://openreview.net/forum?id=18gC6pZVVc}
}

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

d4rt-0.0.4.tar.gz (7.0 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

d4rt-0.0.4-py3-none-any.whl (6.1 kB view details)

Uploaded Python 3

File details

Details for the file d4rt-0.0.4.tar.gz.

File metadata

  • Download URL: d4rt-0.0.4.tar.gz
  • Upload date:
  • Size: 7.0 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: uv/0.8.17

File hashes

Hashes for d4rt-0.0.4.tar.gz
Algorithm Hash digest
SHA256 cb959a902bf2bc996f3e1f946a25258198b454479df203872e0c0a0de41ee9ec
MD5 a7d7fc9ed2e2ad250bceb85dd2a16836
BLAKE2b-256 52dc63af5428ab89a51c85895933e877833845dc8e5d16311e954003905900e7

See more details on using hashes here.

File details

Details for the file d4rt-0.0.4-py3-none-any.whl.

File metadata

  • Download URL: d4rt-0.0.4-py3-none-any.whl
  • Upload date:
  • Size: 6.1 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: uv/0.8.17

File hashes

Hashes for d4rt-0.0.4-py3-none-any.whl
Algorithm Hash digest
SHA256 46fdf1279a5af9b30ed24eb64179f546887d39c0f7f73ab625412fa1528d992e
MD5 9edbbcb2d06744ac742f6ac31c952c3d
BLAKE2b-256 2367af5f4e8f2788af643ca7cb906c2ca9b9ccdddd133c8090b2300e7aaa053b

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page