Datasets:

KangLiao
/

DL3DV-Absolute-Camera

roll float64	pitch float64	vfov float64	parse_ok bool
0.0044	0.1566	1.0864	true
0.014	0.152	1.106	true
0.0164	0.1564	1.116	true
0.034	0.146	1.1046	true
0.04	0.14	1.0866	true
0.0446	0.146	1.0806	true
0.0464	0.1344	1.066	true
0.046	0.144	1.0744	true
0.0406	0.144	1.1066	true
0.0406	0.1466	1.0844	true
0.0346	0.1506	1.1064	true
0.0308	0.154	1.104	true
0.0264	0.15	1.0946	true
0.0206	0.1546	1.086	true
0.016	0.146	1.1004	true
0.0144	0.146	1.0804	true
0.014	0.142	1.076	true
0.0146	0.14	1.0664	true
0.0104	0.1344	1.046	true
0.0044	0.1326	1.0466	true
0.0044	0.1344	1.0764	true
0.0044	0.1346	1.0664	true
0.0004	0.1346	1.0764	true
-0.0004	0.1366	1.0566	true
-0.0004	0.14	1.0566	true
-0.0007	0.136	1.0604	true
-0.004	0.144	1.0564	true
-0.0044	0.1504	1.0506	true
-0.0104	0.154	1.06	true
-0.0106	0.152	1.0404	true
-0.0044	0.1504	1.0446	true
-0.006	0.1504	1.044	true
-0.004	0.1464	1.0604	true
-0.0016	0.1504	1.0406	true
-0.0044	0.146	1.044	true
0.0004	0.144	1.054	true
-0.0044	0.1564	1.0566	true
0.0004	0.146	1.066	true
0.0004	0.15	1.0446	true
0.0044	0.1464	1.0406	true
0.0054	0.1466	1.0466	true
0.0064	0.1466	1.0264	true
0.0104	0.14	1.0246	true
0.0046	0.14	1.0506	true
0.0044	0.1406	1.056	true
0.0064	0.14	1.056	true
0.0064	0.1546	1.0404	true
0.0014	0.145	1.026	true
0.004	0.14	1.006	true
0.0046	0.1406	1.0064	true
0.0044	0.14	1.0064	true
0.0144	0.14	0.9844	true
0.014	0.146	1.0044	true
0.0164	0.14	1.0066	true
0.0166	0.142	0.9744	true
0.0146	0.142	1.0246	true
0.0116	0.146	0.986	true
0.0046	0.144	1.04	true
0.0066	0.1406	1.0406	true
0.004	0.14	1.0404	true
0.0007	0.14	1.016	true
0.0017	0.14	1.0064	true
-0.0044	0.134	1.0464	true
-0.0064	0.14	1.0166	true
-0.014	0.14	1.0406	true
-0.0218	0.14	1.0404	true
-0.0264	0.14	1.0466	true
-0.0308	0.14	1.046	true
-0.034	0.14	1.04	true
-0.0344	0.142	1.0444	true
-0.0304	0.14	1.0404	true
-0.026	0.1408	1.0466	true
-0.024	0.142	1.0506	true
-0.0244	0.154	1.04	true
-0.0204	0.1546	1.0564	true
-0.0206	0.1646	1.018	true
-0.0164	0.166	1.0264	true
-0.014	0.16	1.0404	true
-0.0118	0.16	1.0406	true
-0.0064	0.16	1.04	true
-0.0044	0.16	1.0084	true
0.0017	0.16	1.026	true
0.0017	0.156	1.0064	true
0.0064	0.154	1.04	true
0.0004	0.154	0.9866	true
-0.004	0.1466	0.9766	true
-0.006	0.1366	1.006	true
-0.0104	0.136	1.0044	true
-0.0104	0.126	1.04	true
-0.0164	0.1266	1.0044	true
-0.0144	0.1264	1.0064	true
-0.0204	0.126	0.9646	true
-0.0208	0.1266	0.9766	true
-0.0264	0.1266	0.9604	true
-0.0246	0.115	0.9766	true
-0.0308	0.12	1.006	true
-0.0304	0.115	0.9864	true
-0.0266	0.124	1.0066	true
-0.0264	0.12	0.9664	true
-0.0246	0.125	1.004	true

End of preview.

DL3DV-Absolute-Camera

Per-frame camera parameter annotations for the DL3DV dataset (6,377 scenes across 7 buckets 1K–7K; 2,161,003 valid per-frame annotations), predicted by the Puffin camera-centric multimodal model (Qwen2.5-7B + C-RADIOv3-H visual encoder).

Format

The archive mirrors the source DL3DV-ALL-960P layout: one .zip per scene, grouped into bucket folders — <bucket>/<scene_hash>.zip (e.g. 1K/001dccbc…740d5f.zip). The scene hashes match DL3DV-ALL-960P, so the camera annotations pair 1:1 with the source frames.

Each zip unpacks to the original per-frame layout:

dense/camera/frame_00001.json
dense/camera/frame_00002.json
...

Each JSON holds the predicted monocular camera parameters for that frame:

field	meaning	unit
`roll`	camera roll	radians
`pitch`	camera pitch	radians
`vfov`	vertical field of view	radians
`k1`	radial distortion coefficient	–
`parse_ok`	whether the model output parsed within valid ranges	bool

Example (dense/camera/frame_00001.json):

{"roll": 0.0044, "pitch": 0.1566, "vfov": 1.0864, "k1": 0.0, "parse_ok": true}

Camera-parameter distributions

Histograms of the predicted roll / pitch / vertical-FoV over all frames (proportion of valid samples per 10° bin; parse_ok=False samples excluded).

split	roll μ / med / σ	pitch μ / med / σ	FoV μ / med / σ
all (2.16M frames)	−0.3° / −0.3° / 4.9°	−7.0° / −6.1° / 16.0°	53.4° / 54.8° / 8.7°

Reading the distributions

Roll is tightly peaked at 0° (σ ≈ 4.9°) — DL3DV captures are shot level.
Pitch carries a downward bias (median ≈ −6°, mean ≈ −7°) with a wide spread (σ ≈ 16°), reflecting the varied up/down camera motion across scenes.
FoV concentrates around 50–55° (median ≈ 55°) — moderately wide video capture, ranging ~20–90°.

Notes

Annotations are model predictions, not ground truth.
Generated with greedy decoding (deterministic) at 640px input.
One .zip per scene keeps the repo file count manageable (6,377 archives) while preserving the original dense/camera/frame_*.json structure on unpack.
Figure reproduces scripts/annotation/stat_camera_captions.py from the Puffin repo.

Downloads last month: 1,357

Total file size:

431 MB