Datasets:

Dhi-Technologies
/

fixed-camera-3d-benchmark

Size:

Tags:

License:

Dataset card Files Files and versions

xet

Community

Acknowledge the terms to access this dataset

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

Access is granted for non-commercial research and evaluation use only. By requesting access you agree not to redistribute this dataset or any derivative data, to cite Dhi Technologies in any publication or output that uses it, and to obtain a separate commercial license via dhi-tech.com before any commercial use. Access requests are reviewed manually by Dhi Technologies.

Fixed-Camera 3D Benchmark

Product: fixed-camera-3d ("fixedcam3d"): turns any fixed camera into a metric 3D sensor via self-calibration from ordinary walking people, then reports metric position, speed, and height. Pure Python (numpy/scipy), no GPU, no model weights.

Synthetic data, real-data validation in progress

Scenes are procedurally generated from a known pinhole camera (height, tilt, and focal all ground truth) observing walkers of known height, position, and speed, with detector-like box noise added. The self-calibration and localization pipeline is the real product code run against this synthetic input; nothing about the estimator is mocked. Real annotated CCTV footage has not been run through this pipeline yet, and public multi-camera ground-truth sets such as WILDTRACK have not been evaluated either; that is explicitly the next (GPU-phase) step per the repo's own docs.

What's in this dataset

scenes.jsonl (3 rows, about 74 KB): one row per camera configuration in the default sweep (3 m / 20 degree tilt, 5 m / 30 degree tilt, 8 m / 45 degree tilt; height, tilt, and focal all ground truth). Each row has:
- camera_truth: the ground-truth camera geometry (height_m, tilt_deg, focal_px)
- calibration_observations: 80 synthetic noisy person-detection boxes used to self-calibrate (each with true_xy_m / true_height_m ground truth)
- eval_observations: 60 held-out synthetic boxes used to score the calibrated pipeline
bench_results.json (about 2.3 KB): the fitted camera parameters (recovered height, tilt, focal), the calibration residual, the sensitivity_m_per_half_deg conditioning probe, and the accuracy report for all 3 configs. This is the source of the table below.

How to load it

Verified against the actual files in this repository (prints 3 scenes, the three dict keys, and the position RMSE for the first config):

import json
from huggingface_hub import hf_hub_download

repo_id = "Dhi-Technologies/fixed-camera-3d-benchmark"
scenes_path = hf_hub_download(repo_id, "scenes.jsonl", repo_type="dataset")
bench_path = hf_hub_download(repo_id, "bench_results.json", repo_type="dataset")

scenes = [json.loads(line) for line in open(scenes_path)]
bench = json.load(open(bench_path))

print(len(scenes), "scenes")
print(list(scenes[0].keys()))
print(bench["sweep"][0]["position_rmse_m"])

Measured result (from this repo, `fixedcam3d.cli bench --seed 0`)

Recovered tilt error is the absolute difference between the ground-truth tilt_deg and the self-calibrated tilt_deg, read directly from bench_results.json.

camera (truth)	position RMSE	position p90	speed MAE	height MAE	recovered tilt error	sensitivity (m / half deg)
3 m, 20 deg tilt	4.30 m	5.77 m	0.353 m/s	0.038 m	3.14 deg	1.72
5 m, 30 deg tilt	0.16 m	0.23 m	0.083 m/s	0.056 m	0.43 deg	0.72
8 m, 45 deg tilt	0.19 m	0.22 m	0.076 m/s	0.078 m	0.02 deg	0.20

The 3 m / 20 degree config is disclosed as the hard case: shallow-angle geometry is genuinely ill-conditioned for single-camera self-calibration. The calibrator's own notes field for that row flags it directly (quoted here with the dash replaced by a comma per house style): "shallow/ill-conditioned geometry, low residual but far-field distances are fragile". The optimizer still converges with a low residual (0.037 in this run) but the recovered tilt is off by a few degrees, and at that geometry a few degrees of tilt error blows up into meters of far-field position error. sensitivity_m_per_half_deg is the conditioning probe built to catch exactly this: it perturbs the recovered tilt by half a degree and measures how far the median ground position moves, 1.72 m at the shallow geometry versus 0.72 m and 0.20 m at the two well-conditioned geometries.

Reproduce with: PYTHONPATH=src .venv/bin/python -m fixedcam3d.cli bench --seed 0

Schema notes

Positions are metric (meters), on the ground plane, with the camera as origin unless noted in camera_truth.
The camera model is a pure pinhole with no lens distortion, yaw, or roll terms, and no rolling shutter model.
This is an evaluation-only population: calibration_observations and eval_observations are disjoint synthetic draws for the same ground-truth camera, not a train/test split over real footage.

Method card, no trained weights

Pure Python (numpy/scipy), no GPU and no model weights. Self-calibration solves for camera height, tilt, and focal length from ordinary walking-person detections (a coarse grid search over the three parameters followed by soft-L1 refinement); localization back-projects foot points to the ground plane. The shallow-angle (3 m / 20 degree) config is disclosed as the hard, ill-conditioned case in the table above rather than dropped from it.

Limitations

Synthetic only in this dataset: every scene comes from a known, simulated pinhole camera. No real annotated CCTV footage or public dataset such as WILDTRACK has been run through this pipeline yet.
Single generation seed (seed=0): the table is a point estimate for one sweep, not averaged across multiple seeds.
Ground-contact detection is naive (bottom-center of the detection box treated as the foot point), so occlusion at the frame edge or by other objects can silently produce a wrong ground point, a limitation this synthetic benchmark does not stress because its boxes are clean by construction.
The height prior is a single population-mean scalar per scene, not per-subject, so a systematically biased population (for example, mostly children) would bias calibration in a way this synthetic sweep, which uses a fixed height distribution, does not surface.
Rolling shutter and lens distortion are not modeled, so fast subjects or wide-angle lenses on real hardware would violate assumptions this synthetic data does not test.

License

This dataset is released under CC BY-NC 4.0 (non-commercial). Access is gated and requires manual approval: it is provided for non-commercial research and evaluation only, redistribution is not permitted, and any publication or output using it should cite Dhi Technologies. Commercial use requires a separate agreement; contact dhi-tech.com.