Download web/data/test.json from devmandan/qr-scan-difficulty-check: direct link, hf CLI and curl.
- Browser
- Download file 13.6 kB
-
https://huggingface.co/spaces/devmandan/qr-scan-difficulty-check/resolve/main/web/data/test.json
- Command line
-
hf download hf://spaces/devmandan/qr-scan-difficulty-check/web/data/test.json
-
curl -L -o test.json https://huggingface.co/spaces/devmandan/qr-scan-difficulty-check/resolve/main/web/data/test.json
13.6 kB
| { | |
| "about": "The design test's named choices, the figures behind its \"How this was tested\", and the working check's two pictures, each with its source. The page (test.js, scanner.js) and the server's design test (design.py) both read this file; the server reads it once at start-up, and a missing or broken file turns the test off.", | |
| "version": "2026-11-15", | |
| "page": true, | |
| "scenes": { | |
| "s": 20, | |
| "seeds": [ | |
| 0, | |
| 1, | |
| 2, | |
| 3, | |
| 4, | |
| 5, | |
| 6, | |
| 7, | |
| 8, | |
| 9, | |
| 10, | |
| 11, | |
| 12, | |
| 13, | |
| 14, | |
| 15, | |
| 16, | |
| 17, | |
| 18, | |
| 19 | |
| ], | |
| "streams": "grid:everyday", | |
| "source": "the study's seeds: seeds 0 to 19 of the size table's own everyday streams for the version; 19 is the fewest scenes at which a clean pass rules out a 15% loss (13.9% at 20)" | |
| }, | |
| "step": { | |
| "at": "p(v)", | |
| "source": "one step, at the measured table's need for the version (sizes.json needs); a version between measured ones takes the next one up, version 1 takes version 2's" | |
| }, | |
| "rule": { | |
| "alpha": 0.05, | |
| "like_cost": 2, | |
| "q": 16, | |
| "least_worse": 5, | |
| "source": "worse: a one-sided exact sign test under 0.05 (a definition), which needs a cost of at least 5; as well as a plain code: a cost of 2 or less, the study's flag being 3, with the plain code or the design read in at least 16 of 20 scenes (a named power choice, the same answers from 14 to 16 on the tuning half); otherwise can't tell: close with a cost of 3 or more, else few" | |
| }, | |
| "limits": { | |
| "min_ppm": 2, | |
| "min_ppm_v1": 3, | |
| "max_version": 21, | |
| "max_version_after_b": 21, | |
| "max_pixels": 16000000, | |
| "source": "at least 2 pixels per square in the file (uec_min_ppm in sizes.json; the study's 2-pixel files were tested), 3 for version 1 (uec_min_ppm_v1, the same floor); versions up to 10, and up to 21, the size table's last measured version, once plain codes at 13, 15, 18 and 21 never read worse; at most 16 megapixels, the page's own conversion cap" | |
| }, | |
| "server": { | |
| "deadline_s": 40, | |
| "nice": 10, | |
| "slot_hold_s": 10, | |
| "ticket_extra_s": 5, | |
| "idle_s": 60, | |
| "after_deadline_s": 60, | |
| "scene_wait_s": 10, | |
| "retry_s": [ | |
| 2, | |
| 40 | |
| ], | |
| "max_tickets": 256, | |
| "held_mb_max": 128, | |
| "busy_while_readings_wait": false, | |
| "source": "the server reads these at start-up: its deadline equals the reading's 40 s; the worker's priority, the deadline and the browser's budget are set again from the live timing run; a slot gives its upload 10 s to start; a ticket is held for at most its last Retry-After plus 5 s; a test is dropped after 60 s with no request, or 60 s after its deadline; a scene not yet rendered is waited for 10 s; Retry-After runs from 2 to 40 s; at most 256 tickets wait at once; the scenes held for pages are bounded at 128 MB; the busy answer while readings queue stays off unless the live run calls for it" | |
| }, | |
| "page_waits": { | |
| "settle_s": 1, | |
| "between_tests_s": 2, | |
| "poll_s": 1, | |
| "busy_total_s": 120, | |
| "check_retry_s": 2, | |
| "probe_wait_s": 3, | |
| "scene_retry_s": 1, | |
| "upload_s": 100, | |
| "source": "a choice still standing after 1 s starts a test (the keyboard passes choices it doesn't keep); 2 s between a page's tests as a courtesy; polls about once a second; busy for up to 2 minutes; 2 s before the working check's retry; a test waits at most 3 s for the probe (a cold detect took 0.97 s); a scene's facts are asked for again 1 s after the server says it isn't rendered yet; a slot request unanswered within the slot's own 10 s hold (slot_hold_s), or an upload unanswered after 100 s, ends as the server unreachable, with Try again: the server answers a slot request at once, gives a picture 60 s to arrive, as for a reading, and starts its test within the 40 s deadline" | |
| }, | |
| "budget": { | |
| "fetch_s": 45, | |
| "project_after_scenes": 2, | |
| "detect_s": 5, | |
| "fetch_tries": 2, | |
| "source": "45 s of fetching the scenes' pictures from the test's start, projected from the link's own pace after 2 scenes and after each one from then on; it holds down to about 2 Mbit/s up to version 10 and 3 Mbit/s at 21; a failed fetch is tried once more; each detect has 5 s" | |
| }, | |
| "join": 3, | |
| "join_source": "the browser's scanner joins a page session's tests by itself until it has done so in 3 tests, then is offered with one tap, and Try again keeps it in a test it was to join: usually about 24 MB of test pictures in all (3 x 7.93 MB, the median test with JPEG frames in mb)", | |
| "browsers": { | |
| "brand": "Google Chrome", | |
| "platforms": { | |
| "macOS": "apple", | |
| "Android": "google" | |
| }, | |
| "source": "Google Chrome on macOS and on Android, named by platform after the working check; other browsers get the open-source answer until each is measured" | |
| }, | |
| "frames": { | |
| "capture": [ | |
| 1920, | |
| 1440 | |
| ], | |
| "frame": [ | |
| 1280, | |
| 960 | |
| ], | |
| "jpeg_quality": 82, | |
| "png_level": 1, | |
| "margin_squares": 6, | |
| "twin_margin": 4, | |
| "source": "the size table's everyday scene: each capture a baseline JPEG at quality 82, shrunk to the 1280 x 960 grey frame the scanners read by the whole-number rule (4a + 2b + 2c + d + 4) div 9; the patches and the fallback frames are grey PNGs at compression level 1; up to 6 squares of the design show round its code, and the plain code has a 4-square margin" | |
| }, | |
| "mb": { | |
| "jpeg": { | |
| "10": 9, | |
| "21": 16 | |
| }, | |
| "png": { | |
| "10": 16, | |
| "21": 23 | |
| }, | |
| "source": "the most measured for the code's range, rounded up: with JPEG frames 8.46 MB up to version 10 and 15.02 at version 21; with PNG frames 14.55 and 21.33, plus room for two scenes' JPEGs fetched before a switch; a test usually downloads about 8 MB: 7.93 MB at the median with JPEG frames, over the study's 148 distinct dark-on-light designs at versions 2 to 10" | |
| }, | |
| "check": { | |
| "png": { | |
| "b64": "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", | |
| "w": 148, | |
| "h": 148, | |
| "grey_sha256": "75617964609cd3d236f991324d72afe13f686203826f150753756fc5ccbdba0b", | |
| "content_sha256": "c127abccc1678a8782bbe8aa087f9b68ec1b0f451b3f4322198c9df89e26d965" | |
| }, | |
| "jpeg": { | |
| "b64": "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", | |
| "w": 128, | |
| "h": 96, | |
| "grey_sha256": "d401cd2e47bb5934a59db3b51d93b05e0165a69e3ad0cf56eb122e5a8173575f" | |
| }, | |
| "source": "a plain code of a test link on example.com (version 2, error correction M, mask 2, 4 pixels a square, a 6-square margin) and a small grey JPEG made as the captures are (baseline, quality 82, one channel); each hash is the SHA-256 of its grey bytes as OpenCV decodes it, which is libjpeg-turbo's accurate integer IDCT for the JPEG; the fast and float IDCTs give other hashes" | |
| }, | |
| "figures": { | |
| "N": 125, | |
| "cover": { | |
| "open": { | |
| "y": 53, | |
| "x": 14, | |
| "z": 11 | |
| }, | |
| "open_worse": { | |
| "y": 69, | |
| "x": 64 | |
| }, | |
| "apple": { | |
| "y": 32, | |
| "x": 1, | |
| "z": 3 | |
| }, | |
| "google": { | |
| "y": 38, | |
| "x": 2, | |
| "z": 8 | |
| } | |
| }, | |
| "chrome": { | |
| "apple": { | |
| "os": "14.3.1", | |
| "n": 13040, | |
| "chrome": 157 | |
| }, | |
| "google": { | |
| "os": "14", | |
| "n": 3200, | |
| "chrome": 113, | |
| "play_services": "26.41.31" | |
| } | |
| }, | |
| "limits": { | |
| "big": [ | |
| 3, | |
| 10 | |
| ], | |
| "big_pass": [ | |
| 1, | |
| 30 | |
| ], | |
| "small": [ | |
| 1, | |
| 10 | |
| ], | |
| "small_pass": [ | |
| 2, | |
| 3 | |
| ], | |
| "binomial": [ | |
| 0.035, | |
| 0.677 | |
| ] | |
| }, | |
| "scope": { | |
| "styles": 4, | |
| "versions": [ | |
| 3, | |
| 8 | |
| ] | |
| }, | |
| "load_test": { | |
| "median_x": 1.52, | |
| "readings": 1297, | |
| "one_at_a_time_x": 1.03, | |
| "source": "a load test on this Space, 17 Nov 2026: readings made wholly while a test ran, with both reading slots busy, took about 1.5 times their median time without tests (1.52 times, over 1,297 readings); one reading at a time while tests ran, 1.03 times" | |
| }, | |
| "server_frames": { | |
| "designs": 38, | |
| "frames": 1520, | |
| "differ": 1211, | |
| "same_answer": { | |
| "zxing-cpp": 38, | |
| "wechat": 38, | |
| "apple": 38, | |
| "google": 35 | |
| }, | |
| "cant_tell": 3, | |
| "source": "the study's captures were made on a Mac; for 38 of the study's designs, 1,211 of their 1,520 frames came out different in some pixels on this server; zxing-cpp, WeChat's scanner and Apple's gave the study's answer for all 38, Google's for 35, and for the other 3 it moved to can't tell" | |
| }, | |
| "source": "a study of 163 files from one code generator, each beside a plain code with the same content, in 13,040 simulated captures (20 scenes at the need and at 1.6 times it, for each file and its plain code); its coverage counts are for its 125 distinct styled designs that are dark on light, at the need (of the 163 files, 3 are copies of others, 10 are light on dark and 25 are plain codes), and rest on Apple's and Google's scanners only; Chrome 157 on macOS 14.3.1 read all 13,040 captures exactly as Apple's scanner did, and Chrome 113 on Google's Android 14 emulator with Google Play (Play services 26.41.31) all 3,200 of a 40-design sample exactly as Google's; the limits, each [a, b] for a in b, are the binomial chance that a design losing that share of 20 scans still passes" | |
| }, | |
| "acceptance": { | |
| "band": 24, | |
| "measure_median_x": 1.25, | |
| "measure_p95_x": 1.5, | |
| "source": "set before the live run: the open-source answers allowed to move to or from can't tell on the server's own processor, the number of the held-out half's answers that one changed read (of the 40 per scanner per design) would move to or from can't tell on a Mac, 24 of 170, none of which one read could swap between as well as and worse; and /api/measure's budgets while tests run; no code reads these" | |
| } | |
| } | |