From install to first measurement.

CircleSFR targets Python 3.10+ · NumPy, SciPy, Pillow · macOS, Windows, Linux. The PyPI release is in preparation; until it publishes, evaluation installs are delivered as a wheel by email with your key.

$ pip install circlesfr
$ export CIRCLESFR_LICENSE=/path/to/license.key
# or place the key at ~/.circlesfr/license.key

Keys are Ed25519-signed files that verify offline — nothing phones home, and air-gapped machines work. Evaluation keys run 30 days with watermarked outputs. Request one by email.

Frame the chart to fill the image width, square to the camera. What matters is the field of view the chart covers, not the distance: distance to fill the frame ≈ chart width ÷ (2 × tan(HFOV/2)). Light evenly, expose so neither the white field nor the dark rings clip — the software reports per-channel clipping fractions, so a bad exposure is caught on the first run.

Chart sizes and single-capture ratings: CircleSFR 400 (400 mm, ≤3 MP) · CircleSFR 800 (800 mm, ≤12 MP) · CircleSFR 1200 (1200 mm, ≤28 MP). See the size picker for a recommendation.

>>> import circlesfr
>>> result = circlesfr.analyze("capture_042.tif")
>>> result.summary["sfr50_sag_l"]      # center SFR50, cy/px
0.2609
>>> result.clipping.high               # per-channel clipping
(0.0021, 0.0018, 0.0016)

The package linearizes the capture, registers all 83 ring markers, fits an ellipse to each ring, and measures four edge sectors per ring — sagittal left/right and meridional near/far — giving a dense MTF map across the field from a single image.

Results are Python objects today; CSV export, PDF reports, and a batch CLI are on the roadmap and will be announced when they ship. Frequencies are cycles/pixel with Nyquist at 0.50.