GEN3C
Collection
3D-Informed World-Consistent Video Generation with Precise Camera Control
•
2 items
•
Updated
Dataset Viewer
Search is not available for this dataset
original_shapes
dict | fps
int64 24
24
| data_ranges
dict |
|---|---|---|
{
"image": [
121,
3,
704,
1280
],
"depth": [
121,
1,
704,
1280
],
"mask": [
121,
1,
704,
1280
],
"w2c": [
121,
4,
4
],
"intrinsics": [
121,
3,
3
]
} | 24 | {
"image": [
-1,
1
],
"depth": [
6.4765625,
15.0078125
],
"mask": [
1,
1
],
"w2c": [
-3.1718783378601074,
1
],
"intrinsics": [
0,
1006.3212280273438
]
} |
{
"image": [
121,
3,
704,
1280
],
"depth": [
121,
1,
704,
1280
],
"mask": [
121,
1,
704,
1280
],
"w2c": [
121,
4,
4
],
"intrinsics": [
121,
3,
3
]
} | 24 | {
"image": [
-1,
1
],
"depth": [
6.0546875,
40.875
],
"mask": [
1,
1
],
"w2c": [
-0.05187155306339264,
12.911611557006836
],
"intrinsics": [
0,
640
]
} |
{
"image": [
121,
3,
704,
1280
],
"depth": [
121,
1,
704,
1280
],
"mask": [
121,
1,
704,
1280
],
"w2c": [
121,
4,
4
],
"intrinsics": [
121,
3,
3
]
} | 24 | {
"image": [
-1,
1.000000238418579
],
"depth": [
0.7930031418800354,
2.072265625
],
"mask": [
1,
1
],
"w2c": [
-0.06698334217071533,
1
],
"intrinsics": [
0,
1634.5882568359375
]
} |
{
"image": [
121,
3,
704,
1280
],
"depth": [
121,
1,
704,
1280
],
"mask": [
121,
1,
704,
1280
],
"w2c": [
121,
4,
4
],
"intrinsics": [
121,
3,
3
]
} | 24 | {
"image": [
-1,
1
],
"depth": [
1.326171875,
6.37109375
],
"mask": [
1,
1
],
"w2c": [
-0.16525442898273468,
1
],
"intrinsics": [
0,
1446.695556640625
]
} |
{
"image": [
121,
3,
704,
1280
],
"depth": [
121,
1,
704,
1280
],
"mask": [
121,
1,
704,
1280
],
"w2c": [
121,
4,
4
],
"intrinsics": [
121,
3,
3
]
} | 24 | {
"image": [
-1,
1.000000238418579
],
"depth": [
1.3998546600341797,
3.445312976837158
],
"mask": [
1,
1
],
"w2c": [
-0.1701044887304306,
5.418583869934082
],
"intrinsics": [
0,
2968.870849609375
]
} |
{
"image": [
121,
3,
704,
1280
],
"depth": [
121,
1,
704,
1280
],
"mask": [
121,
1,
704,
1280
],
"w2c": [
121,
4,
4
],
"intrinsics": [
121,
3,
3
]
} | 24 | {
"image": [
-1,
1
],
"depth": [
1.949218511581421,
2.14453125
],
"mask": [
1,
1
],
"w2c": [
-0.6437628865242004,
1
],
"intrinsics": [
0,
2194.381103515625
]
} |
{
"image": [
121,
3,
704,
1280
],
"depth": [
121,
1,
704,
1280
],
"mask": [
121,
1,
704,
1280
],
"w2c": [
121,
4,
4
],
"intrinsics": [
121,
3,
3
]
} | 24 | {
"image": [
-1,
1
],
"depth": [
2.482421875,
2.599609375
],
"mask": [
1,
1
],
"w2c": [
-0.3528679311275482,
1
],
"intrinsics": [
0,
640
]
} |
{
"image": [
121,
3,
704,
1280
],
"depth": [
121,
1,
704,
1280
],
"mask": [
121,
1,
704,
1280
],
"w2c": [
121,
4,
4
],
"intrinsics": [
121,
3,
3
]
} | 24 | {
"image": [
-1,
1
],
"depth": [
2.087890625,
4.484375
],
"mask": [
1,
1
],
"w2c": [
-0.32845884561538696,
1
],
"intrinsics": [
0,
2241.742431640625
]
} |
{
"image": [
121,
3,
704,
1280
],
"depth": [
121,
1,
704,
1280
],
"mask": [
121,
1,
704,
1280
],
"w2c": [
121,
4,
4
],
"intrinsics": [
121,
3,
3
]
} | 24 | {
"image": [
-1,
1.000000238418579
],
"depth": [
14.53905963897705,
18.979455947875977
],
"mask": [
1,
1
],
"w2c": [
-0.12961345911026,
1.6763273477554321
],
"intrinsics": [
0,
11908.7265625
]
} |
{
"image": [
121,
3,
704,
1280
],
"depth": [
121,
1,
704,
1280
],
"mask": [
121,
1,
704,
1280
],
"w2c": [
121,
4,
4
],
"intrinsics": [
121,
3,
3
]
} | 24 | {
"image": [
-1,
1
],
"depth": [
1.9769967794418335,
5.906250953674316
],
"mask": [
1,
1
],
"w2c": [
-0.0346110463142395,
1
],
"intrinsics": [
0,
1160.22705078125
]
} |
GEN3C: 3D-Informed World-Consistent Video Generation with Precise Camera Control
CVPR 2025 (Highlight)
Xuanchi Ren*,
Tianchang Shen*
Jiahui Huang,
Huan Ling,
Yifan Lu,
Merlin Nimier-David,
Thomas Müller,
Alexander Keller,
Sanja Fidler,
Jun Gao
* indicates equal contribution
Paper, Project Page
Abstract: We present GEN3C, a generative video model with precise Camera Control and temporal 3D Consistency. Prior video models already generate realistic videos, but they tend to leverage little 3D information, leading to inconsistencies, such as objects popping in and out of existence. Camera control, if implemented at all, is imprecise, because camera parameters are mere inputs to the neural network which must then infer how the video depends on the camera. In contrast, GEN3C is guided by a 3D cache: point clouds obtained by predicting the pixel-wise depth of seed images or previously generated frames. When generating the next frames, GEN3C is conditioned on the 2D renderings of the 3D cache with the new camera trajectory provided by the user. Crucially, this means that GEN3C neither has to remember what it previously generated nor does it have to infer the image structure from the camera pose. The model, instead, can focus all its generative power on previously unobserved regions, as well as advancing the scene state to the next frame. Our results demonstrate more precise camera control than prior work, as well as state-of-the-art results in sparse-view novel view synthesis, even in challenging settings such as driving scenes and monocular dynamic video. Results are best viewed in videos.
Citation
@inproceedings{ren2025gen3c,
title={GEN3C: 3D-Informed World-Consistent Video Generation with Precise Camera Control},
author={Ren, Xuanchi and Shen, Tianchang and Huang, Jiahui and Ling, Huan and
Lu, Yifan and Nimier-David, Merlin and Müller, Thomas and Keller, Alexander and
Fidler, Sanja and Gao, Jun},
booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
year={2025}
}
- Downloads last month
- 103
Size of downloaded dataset files:
6.33 kB
Size of the auto-converted Parquet files:
44 kB
Number of rows:
10