feat(tiling): 自适应瓦片规划器 + VLM 分辨率离线标定
背景：逐图框审查固定「整体 + 2x2 象限」截图，大图幅（A0/A1）读不清细节。 改为按**实体分布**自适应决定截几张、截哪里。 - 新增 dwg_review/tiling.py（纯函数）： SHEET_SIZES_MM / frame_world_size / frame_scale（sheet_scale>0 用之，否则由 bbox/纸张反算，否则 default）/ effective_px（与 downsample 同规则）/ world_per_px / is_text_type（标准名 + 自定义类名含 TEXT/LABEL/DIM/NOTE/TAG/ ANNO/TITLE）/ cluster_text_flags / tiles_for_box / subdivide / plan_regions （文字簇优先、空白区不出图、max_tiles 截断、窗口夹到图框范围） - 新增 verify_vlm_resolution_offline.py：PIL 合成类图纸图（精确像素字高 + 可控线间距/线宽/虚线）→ 问 VLM 强约束 JSON → 定 R_text / R_line。 实测：R_text=12px（黑底；白底 10px）；R_line=6px。 另：实测该模型**无法可靠计数**（6 条线在 2~10px 间距下答 4/10/4/5/4/4）， 故线条判据改用「1 条 vs 2 条」二选一。 - config 新增 TILING_ENABLED / TEXT_HEIGHT_MM / TEXT_READABLE_PX=12 / LINE_READABLE_PX=6 / TILE_MAX_COUNT / TILE_OVERLAP / HEATMAP_CELL_DIV / HEATMAP_CLUSTER_DILATE / HEATMAP_MAX_AREA_FRAC；EVIDENCE_DETAIL_TILES 与 EVIDENCE_MAX_IMAGES 保留作回退路径参数 - llm_client 新增 image_size_b64（测量绘图区实际像素，决定单张覆盖范围） - phases/evidence.capture_images 支持瓦片规划：先截整体图并测像素 → plan_regions 成功用规划窗口，否则退回 sub_windows（行为零变化）； 末尾始终复位整体视口。collect_frame 传实体/文字 bbox；任一信封 total>returned（截断）→ 退回固定策略，避免误删有内容的瓦片 - 测试：tests/test_tiling.py（28 项）、tests/test_evidence.py 新增规划路径与 降级用例、test_config / test_llm_client 补新常量与 image_size_b64； 全量 283 项全绿
This commit is contained in:
gjm committed
1 parent
9b8a2329df
commit
cba05b0c65
9 files changed
+1246
-10
No files matched your search
+151
-1
@@ -1,8 +1,17 @@
|
||||
"""`dwg_review.phases.evidence` 单元测试。"""
|
||||
|
||||
import base64
|
||||
import io
|
||||
|
||||
import pytest
|
||||
|
||||
from dwg_review.phases.evidence import bbox_to_window, collect_frame
|
||||
from dwg_review.phases.evidence import (
|
||||
_is_truncated,
|
||||
_plan_tiles,
|
||||
_text_boxes,
|
||||
bbox_to_window,
|
||||
collect_frame,
|
||||
)
|
||||
from tests.fakes import FakeMcpClient, frame, make_ctx, ok_image, ok_text
|
||||
|
||||
|
||||
@@ -141,3 +150,144 @@ def test_collect_returns_none_for_unsupported_source():
|
||||
|
||||
ctx = make_ctx(client=FakeMcpClient())
|
||||
assert collect(ctx, 1, {"source_type": "pdf", "bbox": [0, 0, 1, 1]}) is None
|
||||
|
||||
|
||||
# ---------------------------------------------------------------------------
|
||||
# 自适应瓦片规划
|
||||
# ---------------------------------------------------------------------------
|
||||
@pytest.fixture
|
||||
def tiling_env(monkeypatch):
|
||||
"""把瓦片相关配置钉死，使用例不受外部环境变量影响。"""
|
||||
from dwg_review import config
|
||||
|
||||
for name, value in {
|
||||
"TILING_ENABLED": True,
|
||||
"TEXT_HEIGHT_MM": 2.0,
|
||||
"TEXT_READABLE_PX": 12,
|
||||
"LINE_READABLE_PX": 6,
|
||||
"TILE_MAX_COUNT": 36,
|
||||
"TILE_OVERLAP": 0.0,
|
||||
"HEATMAP_CELL_DIV": 32,
|
||||
"HEATMAP_CLUSTER_DILATE": 0,
|
||||
"HEATMAP_MAX_AREA_FRAC": 0.5,
|
||||
"IMAGE_MAX_EDGE": 1280,
|
||||
"EVIDENCE_MAX_IMAGES": 5,
|
||||
"EVIDENCE_DETAIL_TILES": 4,
|
||||
}.items():
|
||||
monkeypatch.setattr(config, name, value)
|
||||
return config
|
||||
|
||||
|
||||
def _png_b64(w=2000, h=1414):
|
||||
from PIL import Image
|
||||
|
||||
buf = io.BytesIO()
|
||||
Image.new("RGB", (w, h), "black").save(buf, format="PNG")
|
||||
return base64.b64encode(buf.getvalue()).decode("ascii")
|
||||
|
||||
|
||||
def _tiling_client(b64, entity_type="LINE"):
|
||||
return FakeMcpClient(
|
||||
responses={
|
||||
"query_entities": ok_text(
|
||||
{"total": 1, "items": [{"handle": "A", "type": entity_type,
|
||||
"bbox": [0, 0, 60000, 20000]}]}
|
||||
),
|
||||
"get_text_content": ok_text({"total": 0, "items": []}),
|
||||
"get_dimensions": ok_text({"total": 0, "items": []}),
|
||||
"get_block_references": ok_text({"total": 0, "items": []}),
|
||||
"get_polylines_info": ok_text({"total": 0, "items": []}),
|
||||
"get_viewport_screenshot": ok_image(b64),
|
||||
}
|
||||
)
|
||||
|
||||
|
||||
def test_is_truncated():
|
||||
assert _is_truncated({"total": 5, "items": [1]}) is True
|
||||
assert _is_truncated({"total": 1, "items": [1]}) is False
|
||||
assert _is_truncated({"items": [1, 2]}) is False
|
||||
assert _is_truncated(None, {}, {"total": 3, "items": []}) is True
|
||||
assert _is_truncated() is False
|
||||
|
||||
|
||||
def test_text_boxes_includes_std_and_custom():
|
||||
entities = {"items": [
|
||||
{"type": "LINE", "bbox": [0, 0, 1, 1]},
|
||||
{"type": "AECDLABEL", "bbox": [2, 2, 3, 3]},
|
||||
{"type": "TEXT", "bbox": [4, 4, 5, 5]},
|
||||
]}
|
||||
texts = {"items": [{"bbox": [6, 6, 7, 7]}]}
|
||||
dimensions = {"items": [{"bbox": [8, 8, 9, 9]}]}
|
||||
boxes = _text_boxes(entities, texts, dimensions)
|
||||
for expected in ((2, 2, 3, 3), (4, 4, 5, 5), (6, 6, 7, 7), (8, 8, 9, 9)):
|
||||
assert tuple(float(v) for v in expected) in boxes
|
||||
assert (0.0, 0.0, 1.0, 1.0) not in boxes # 普通线不算文字
|
||||
|
||||
|
||||
def test_plan_tiles_degrades(tiling_env):
|
||||
win = {"min_x": 0.0, "min_y": 0.0, "max_x": 100.0, "max_y": 50.0}
|
||||
f = frame([0, 0, 100, 50], sheet_size="A0", sheet_scale=100.0)
|
||||
boxes = [(0.0, 0.0, 10.0, 10.0)]
|
||||
assert _plan_tiles(win, f, boxes, [], False, (100, 50)) != [] # 正常路径非空
|
||||
assert _plan_tiles(win, None, boxes, [], False, (100, 50)) == [] # 缺图框
|
||||
assert _plan_tiles(win, f, [], [], False, (100, 50)) == [] # 无实体
|
||||
assert _plan_tiles(win, f, boxes, [], False, None) == [] # 无像素
|
||||
assert _plan_tiles(win, f, boxes, [], True, (100, 50)) == [] # 数据截断
|
||||
# 比例未知（sheet_scale=0 且 sheet_size 不在纸张表）→ 无法规划
|
||||
unknown = frame([0, 0, 100, 50], sheet_size="ZZ", sheet_scale=0.0)
|
||||
assert _plan_tiles(win, unknown, boxes, [], False, (100, 50)) == []
|
||||
|
||||
|
||||
def test_plan_tiles_disabled_by_config(tiling_env, monkeypatch):
|
||||
tiling_env.TILING_ENABLED = False
|
||||
win = {"min_x": 0.0, "min_y": 0.0, "max_x": 100.0, "max_y": 50.0}
|
||||
f = frame([0, 0, 100, 50], sheet_size="A0", sheet_scale=100.0)
|
||||
assert _plan_tiles(win, f, [(0.0, 0.0, 10.0, 10.0)], [], False, (100, 50)) == []
|
||||
|
||||
|
||||
def test_capture_images_tiles_line_cluster(tiling_env):
|
||||
"""线条簇：单张可覆盖 → 2 张瓦片（行优先）。"""
|
||||
pytest.importorskip("PIL")
|
||||
ctx = make_ctx(client=_tiling_client(_png_b64()))
|
||||
ev = collect_frame(
|
||||
ctx, 1, frame([0, 0, 120000, 40000], sheet_size="A0", sheet_scale=100.0)
|
||||
)
|
||||
assert [i["label"] for i in ev["images"]] == ["overall", "tile-1", "tile-2"]
|
||||
|
||||
|
||||
def test_capture_images_tiles_text_cluster_is_finer(tiling_env):
|
||||
"""文字簇用更细尺度 → 瓦片更多，并受 EVIDENCE_MAX_IMAGES 截断。"""
|
||||
pytest.importorskip("PIL")
|
||||
ctx = make_ctx(client=_tiling_client(_png_b64(), entity_type="AECDLABEL"))
|
||||
ev = collect_frame(
|
||||
ctx, 1, frame([0, 0, 120000, 40000], sheet_size="A0", sheet_scale=100.0)
|
||||
)
|
||||
labels = [i["label"] for i in ev["images"]]
|
||||
assert labels == ["overall", "tile-1", "tile-2", "tile-3", "tile-4"] # 上限 5 张
|
||||
|
||||
|
||||
def test_capture_images_falls_back_without_scale(tiling_env):
|
||||
"""缺 sheet_size/sheet_scale → 退回固定 2x2 象限。"""
|
||||
pytest.importorskip("PIL")
|
||||
ctx = make_ctx(client=_tiling_client(_png_b64()))
|
||||
ev = collect_frame(ctx, 1, frame([0, 0, 120000, 40000], sheet_size="ZZ"))
|
||||
assert [i["label"] for i in ev["images"]] == [
|
||||
"overall", "tile-1", "tile-2", "tile-3", "tile-4",
|
||||
]
|
||||
|
||||
|
||||
def test_capture_images_falls_back_when_truncated(tiling_env):
|
||||
"""实体被截断 → 禁用内容过滤，退回固定策略。"""
|
||||
pytest.importorskip("PIL")
|
||||
client = _tiling_client(_png_b64())
|
||||
client.responses["query_entities"] = ok_text(
|
||||
{"total": 500, "items": [{"handle": "A", "type": "LINE",
|
||||
"bbox": [0, 0, 60000, 20000]}]}
|
||||
)
|
||||
ctx = make_ctx(client=client)
|
||||
ev = collect_frame(
|
||||
ctx, 1, frame([0, 0, 120000, 40000], sheet_size="A0", sheet_scale=100.0)
|
||||
)
|
||||
assert [i["label"] for i in ev["images"]] == [
|
||||
"overall", "tile-1", "tile-2", "tile-3", "tile-4",
|
||||
]
|
||||
Reference in new issue
Block a user