FAILURE MAP
← Case archive

FA-78876 / Image orientation metadata / Open access

Face boxes on clockwise-rotated photos land on the mirrored side · case 01

Detected faces on portrait phone shots are highlighted on the opposite side of the frame.

Verified by executionVariant 1 · 8 checks per implementationDownload source bundle ↓JSON ↗

ROOT CAUSE

Tag 6 is decomposed as flip-x then transpose, which is the tag-8 rotation.

THE FAILURE

Tag 6 is decomposed as flip-x then transpose, which is the tag-8 rotation.

Unsuccessful approach: Flipping both axes before transposing is the transverse tag 7, not a clockwise quarter turn.

Case contract

Map a half-open face/crop region [x, y, width, height] given in stored pixel space of an image of size [w, h] into displayed space for orientation tag 1..8. The region is first clipped to the stored image; an empty clipped region returns None, and an unknown tag returns None. Mirroring a half-open interval [a, b) over extent e gives [e-b, e-a); tags 5..8 exchange axes after the flips.

Why this case matters

Camera, phone and scanner images carry an orientation hint separately from the stored pixels; galleries, thumbnailers, editors and upload pipelines must interpret it consistently or photos appear sideways, mirrored or doubly rotated.

1 / The failure

Exit 1
"""Failure Map reference implementation. Python standard library only."""
import json

N = 1
observations = []
def solve(size, tag, box):
    w, h = size
    x0, y0, bw, bh = box
    x1 = min(x0 + bw, w)
    y1 = min(y0 + bh, h)
    x0 = max(x0, 0)
    y0 = max(y0, 0)
    if x1 <= x0 or y1 <= y0:
        return None
    table = {1: (0, 0, 0), 2: (1, 0, 0), 3: (1, 1, 0), 4: (0, 1, 0), 5: (0, 0, 1), 6: (1, 0, 1), 7: (1, 1, 1), 8: (1, 0, 1)}
    if tag not in table:
        return None
    fx, fy, tr = table[tag]
    if fx:
        x0, x1 = w - x1, w - x0
    if fy:
        y0, y1 = h - y1, h - y0
    if tr:
        x0, y0, x1, y1 = y0, x0, y1, x1
    return [x0, y0, x1 - x0, y1 - y0]
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
fixtures = [[[([5, 1], 6, [-3, -2, 5, 4]), [0, 0, 1, 2]], [([3, 7], 6, [-3, 6, 4, 5]), [0, 0, 1, 1]], [([7, 5], 1, [6, -1, 4, 0]), None], [([3, 4], 3, [-2, 3, 0, 2]), None], [([2, 4], 8, [2, 1, 3, 3]), None], [([2, 4], 7, [1, -3, 3, 3]), None], [([2, 6], 6, [4, -1, 1, 4]), None], [([6, 7], 6, [-1, 1, 4, 3]), [3, 0, 3, 3]]], [[([6, 3], 6, [3, 0, 1, 5]), [0, 3, 3, 1]], [([7, 7], 6, [1, 1, 4, 5]), [1, 1, 5, 4]], [([3, 5], 0, [-3, -2, 2, 4]), None], [([3, 1], 1, [2, 2, 2, 0]), None], [([4, 1], 3, [6, 3, 2, 4]), None], [([5, 5], 6, [6, 3, 5, 0]), None], [([4, 5], 1, [1, 2, 5, 4]), [1, 2, 3, 3]], [([6, 6], 6, [4, 4, 3, 3]), [0, 4, 2, 2]]], [[([6, 6], 6, [1, 5, 2, 5]), [0, 1, 1, 2]], [([4, 2], 6, [-2, -2, 5, 4]), [0, 0, 2, 3]], [([1, 7], 6, [3, 6, 1, 4]), None], [([4, 5], 8, [2, -2, 0, 2]), None], [([1, 5], 9, [2, 3, 4, 4]), None], [([6, 1], 2, [2, 1, 3, 0]), None], [([1, 2], 9, [-2, 1, 1, 3]), None], [([5, 4], 6, [4, 3, 3, 3]), [0, 4, 1, 1]]], [[([5, 5], 6, [1, 1, 2, 5]), [0, 1, 4, 2]], [([4, 1], 6, [1, -2, 4, 5]), [0, 1, 1, 3]], [([6, 4], 6, [-1, 3, 2, 2]), [0, 0, 1, 1]], [([3, 6], 8, [6, -2, 2, 3]), None], [([7, 7], 2, [-2, 0, 0, 4]), None], [([2, 5], 1, [1, 6, 5, 2]), None], [([5, 5], 1, [6, 1, 5, 4]), None], [([3, 4], 6, [0, 3, 4, 4]), [0, 0, 1, 3]]], [[([5, 3], 6, [-2, -2, 5, 4]), [1, 0, 2, 3]], [([7, 5], 6, [1, 0, 2, 4]), [1, 1, 4, 2]], [([7, 3], 6, [4, -2, 1, 3]), [2, 4, 1, 1]], [([1, 1], 7, [1, -3, 2, 0]), None], [([4, 5], 6, [1, 3, 0, 5]), None], [([4, 1], 0, [2, 4, 5, 5]), None], [([6, 7], 7, [6, 0, 3, 1]), None], [([7, 7], 6, [0, 5, 4, 3]), [0, 0, 2, 4]]]]
labels = ["regression: tag 6 decomposition entry", "repair trap", "combined fault", "control", "control", "boundary", "boundary", "control"]
for i, (args, expected) in enumerate(fixtures[N-1]):
    check("%s %d" % (labels[i % len(labels)], i), solve(*args), expected)
print(json.dumps({"observations": observations, "passed": all(x["passed"] for x in observations)}, ensure_ascii=False))
raise SystemExit(0 if all(x["passed"] for x in observations) else 1)
Boundary fixtureActualExpectedOutcome
regression: tag 6 decomposition entry 0[0, 3, 1, 2][0, 0, 1, 2]Failed
repair trap 1[6, 2, 1, 1][0, 0, 1, 1]Failed
combined fault 2NoneNonePassed
control 3NoneNonePassed
control 4NoneNonePassed
boundary 5NoneNonePassed
boundary 6NoneNonePassed
control 7[1, 3, 3, 3][3, 0, 3, 3]Failed

SHA-256 / 21ae94380fde01557b9c2a3808b0ab130e87fb9ee8a6971ceb8416ae7efb91df

2 / The unsuccessful fix

Exit 1
"""Failure Map reference implementation. Python standard library only."""
import json

N = 1
observations = []
def solve(size, tag, box):
    w, h = size
    x0, y0, bw, bh = box
    x1 = min(x0 + bw, w)
    y1 = min(y0 + bh, h)
    x0 = max(x0, 0)
    y0 = max(y0, 0)
    if x1 <= x0 or y1 <= y0:
        return None
    table = {1: (0, 0, 0), 2: (1, 0, 0), 3: (1, 1, 0), 4: (0, 1, 0), 5: (0, 0, 1), 6: (1, 1, 1), 7: (1, 1, 1), 8: (1, 0, 1)}
    if tag not in table:
        return None
    fx, fy, tr = table[tag]
    if fx:
        x0, x1 = w - x1, w - x0
    if fy:
        y0, y1 = h - y1, h - y0
    if tr:
        x0, y0, x1, y1 = y0, x0, y1, x1
    return [x0, y0, x1 - x0, y1 - y0]
def check(label, actual, expected):
    observations.append({"check": label, "actual": actual, "expected": expected, "passed": actual == expected})
fixtures = [[[([5, 1], 6, [-3, -2, 5, 4]), [0, 0, 1, 2]], [([3, 7], 6, [-3, 6, 4, 5]), [0, 0, 1, 1]], [([7, 5], 1, [6, -1, 4, 0]), None], [([3, 4], 3, [-2, 3, 0, 2]), None], [([2, 4], 8, [2, 1, 3, 3]), None], [([2, 4], 7, [1, -3, 3, 3]), None], [([2, 6], 6, [4, -1, 1, 4]), None], [([6, 7], 6, [-1, 1, 4, 3]), [3, 0, 3, 3]]], [[([6, 3], 6, [3, 0, 1, 5]), [0, 3, 3, 1]], [([7, 7], 6, [1, 1, 4, 5]), [1, 1, 5, 4]], [([3, 5], 0, [-3, -2, 2, 4]), None], [([3, 1], 1, [2, 2, 2, 0]), None], [([4, 1], 3, [6, 3, 2, 4]), None], [([5, 5], 6, [6, 3, 5, 0]), None], [([4, 5], 1, [1, 2, 5, 4]), [1, 2, 3, 3]], [([6, 6], 6, [4, 4, 3, 3]), [0, 4, 2, 2]]], [[([6, 6], 6, [1, 5, 2, 5]), [0, 1, 1, 2]], [([4, 2], 6, [-2, -2, 5, 4]), [0, 0, 2, 3]], [([1, 7], 6, [3, 6, 1, 4]), None], [([4, 5], 8, [2, -2, 0, 2]), None], [([1, 5], 9, [2, 3, 4, 4]), None], [([6, 1], 2, [2, 1, 3, 0]), None], [([1, 2], 9, [-2, 1, 1, 3]), None], [([5, 4], 6, [4, 3, 3, 3]), [0, 4, 1, 1]]], [[([5, 5], 6, [1, 1, 2, 5]), [0, 1, 4, 2]], [([4, 1], 6, [1, -2, 4, 5]), [0, 1, 1, 3]], [([6, 4], 6, [-1, 3, 2, 2]), [0, 0, 1, 1]], [([3, 6], 8, [6, -2, 2, 3]), None], [([7, 7], 2, [-2, 0, 0, 4]), None], [([2, 5], 1, [1, 6, 5, 2]), None], [([5, 5], 1, [6, 1, 5, 4]), None], [([3, 4], 6, [0, 3, 4, 4]), [0, 0, 1, 3]]], [[([5, 3], 6, [-2, -2, 5, 4]), [1, 0, 2, 3]], [([7, 5], 6, [1, 0, 2, 4]), [1, 1, 4, 2]], [([7, 3], 6, [4, -2, 1, 3]), [2, 4, 1, 1]], [([1, 1], 7, [1, -3, 2, 0]), None], [([4, 5], 6, [1, 3, 0, 5]), None], [([4, 1], 0, [2, 4, 5, 5]), None], [([6, 7], 7, [6, 0, 3, 1]), None], [([7, 7], 6, [0, 5, 4, 3]), [0, 0, 2, 4]]]]
labels = ["regression: tag 6 decomposition entry", "repair trap", "combined fault", "control", "control", "boundary", "boundary", "control"]
for i, (args, expected) in enumerate(fixtures[N-1]):
    check("%s %d" % (labels[i % len(labels)], i), solve(*args), expected)
print(json.dumps({"observations": observations, "passed": all(x["passed"] for x in observations)}, ensure_ascii=False))
raise SystemExit(0 if all(x["passed"] for x in observations) else 1)
Boundary fixtureActualExpectedOutcome
regression: tag 6 decomposition entry 0[0, 3, 1, 2][0, 0, 1, 2]Failed
repair trap 1[0, 2, 1, 1][0, 0, 1, 1]Failed
combined fault 2NoneNonePassed
control 3NoneNonePassed
control 4NoneNonePassed
boundary 5NoneNonePassed
boundary 6NoneNonePassed
control 7[3, 3, 3, 3][3, 0, 3, 3]Failed

SHA-256 / 88d8d526d41882840b56df049184c0ff83014a70ac8dcb67f49cce1047647859

HELD IN THE MEMBER ARCHIVE

The verified repair and its recorded checks are member-only.

This mechanism has 8 recorded checks per implementation. The open-access tier publishes the failure and the unsuccessful fix; the repaired source that passes every check, and the observations that prove it, are available to members.

Every case sharing this mechanism uses the same contract and the same repair, so this one record is held back for all of them.

Member access is invitation-based. Sign in with your invited account to inspect the repair.

Sign in to the archive ↗

Verification & scope

A deterministic bounded teaching model with a stipulated contract; it makes no claim of conformance to any published specification. This reproducer isolates one failure mechanism. Results cover the supplied fixtures. Variants within a family share a test contract and should remain grouped when constructing evaluation splits. Related mechanisms with a shared evaluation_group must also remain together; these controlled models are not independent production incidents.

Observations recorded using Python 3.12.14 at 2026-09-29T14:49:39.247729+00:00.

Case digest / d3362e37d8f62c16eecc1c1b4779ef8cbb26e5fd65ab032b1640879c1804da3e