admin
689a9ba283
- 打通工作区真实标注闭环：支持手工多边形、矩形、圆形、点区域和线段生成 mask，并可保存、回显、更新和删除后端 annotation。 - 增强 polygon 编辑器：支持顶点拖动、顶点删除、边中点插入、多 polygon 子区域选择编辑，以及区域合并和区域去除。 - 接入 GT mask 导入：后端支持二值/多类别 mask 拆分、contour 转 polygon、distance transform seed point，前端支持导入、回显和 seed point 拖动编辑。 - 完善导出能力：COCO JSON 导出对齐前端，PNG mask ZIP 同时包含单标注 mask、按 zIndex 融合的 semantic_frame 和 semantic_classes.json。 - 打通异步任务管理：新增任务取消、重试、失败详情接口与 Dashboard 控件，worker 支持取消状态检查并通过 Redis/WebSocket 推送 cancelled 事件。 - 对接 Dashboard 后端数据：概览统计、解析队列和实时流转记录从 FastAPI 聚合接口与 WebSocket 更新。 - 增强 AI 推理参数：前端发送 crop_to_prompt、auto_filter_background 和 min_score，后端支持点/框 prompt 局部裁剪推理、结果回映射和负向点/低分过滤。 - 接入 SAM3 基础设施：新增独立 Python 3.12 sam3 环境安装脚本、外部 worker helper、后端桥接和真实 Python/CUDA/包/HF checkpoint access 状态检测。 - 保留 SAM3 授权边界：当前官方 facebook/sam3 gated 权重未授权时状态接口会返回不可用，不伪装成可推理。 - 增强前端状态管理：新增 mask undo/redo 历史栈、AI 模型选择状态、保存状态 dirty/draft/saved 流转和项目状态归一化。 - 更新前端 API 封装：补充 annotation CRUD、GT mask import、mask ZIP export、task cancel/retry/detail、AI runtime status 和 prediction options。 - 更新 UI 控件：ToolsPalette、AISegmentation、VideoWorkspace 和 CanvasArea 接入真实操作、导入导出、撤销重做、任务控制和模型状态。 - 新增 polygon-clipping 依赖，用于前端区域 union/difference 几何运算。 - 完善后端 schemas/status/progress：补充 AI 模型外部状态字段、任务 cancelled 状态和进度事件 payload。 - 补充测试覆盖：新增后端任务控制、SAM3 桥接、GT mask、导出融合、AI options 测试；补充前端 Canvas、Dashboard、VideoWorkspace、ToolsPalette、API 和 store 测试。 - 更新 README、AGENTS 和 doc 文档：冻结当前需求/设计/测试计划，标注真实功能、剩余 Mock、SAM3 授权边界和后续实施顺序。
349 lines
12 KiB
Python
349 lines
12 KiB
Python
import numpy as np
|
|
import cv2
|
|
|
|
|
|
def _create_project_and_frame(client):
|
|
project = client.post("/api/projects", json={"name": "AI Project"}).json()
|
|
frame = client.post(f"/api/projects/{project['id']}/frames", json={
|
|
"project_id": project["id"],
|
|
"frame_index": 0,
|
|
"image_url": "frames/0.jpg",
|
|
"width": 640,
|
|
"height": 360,
|
|
}).json()
|
|
template = client.post("/api/templates", json={
|
|
"name": "Template",
|
|
"color": "#06b6d4",
|
|
"z_index": 0,
|
|
"classes": [],
|
|
"rules": [],
|
|
}).json()
|
|
return project, frame, template
|
|
|
|
|
|
def test_predict_accepts_point_object_with_labels(client, monkeypatch):
|
|
_, frame, _ = _create_project_and_frame(client)
|
|
calls = {}
|
|
|
|
monkeypatch.setattr("routers.ai._load_frame_image", lambda frame: np.zeros((10, 10, 3), dtype=np.uint8))
|
|
|
|
def fake_predict_points(image, points, labels):
|
|
calls["args"] = (points, labels)
|
|
return (
|
|
[[[0.1, 0.1], [0.9, 0.1], [0.9, 0.9]]],
|
|
[0.95],
|
|
)
|
|
|
|
monkeypatch.setattr("routers.ai.sam_registry.predict_points", lambda model, image, points, labels: fake_predict_points(image, points, labels))
|
|
|
|
response = client.post("/api/ai/predict", json={
|
|
"image_id": frame["id"],
|
|
"prompt_type": "point",
|
|
"prompt_data": {"points": [[0.5, 0.5], [0.1, 0.1]], "labels": [1, 0]},
|
|
})
|
|
|
|
assert response.status_code == 200
|
|
assert response.json()["scores"] == [0.95]
|
|
assert calls["args"] == ([[0.5, 0.5], [0.1, 0.1]], [1, 0])
|
|
|
|
|
|
def test_predict_applies_crop_and_background_filter_options(client, monkeypatch):
|
|
_, frame, _ = _create_project_and_frame(client)
|
|
calls = {}
|
|
monkeypatch.setattr("routers.ai._load_frame_image", lambda frame: np.zeros((100, 200, 3), dtype=np.uint8))
|
|
|
|
def fake_predict_points(model, image, points, labels):
|
|
calls["shape"] = image.shape
|
|
calls["points"] = points
|
|
calls["labels"] = labels
|
|
return (
|
|
[
|
|
[[0.0, 0.0], [0.2, 0.0], [0.2, 0.2]],
|
|
[[0.45, 0.45], [0.55, 0.45], [0.55, 0.55]],
|
|
],
|
|
[0.9, 0.01],
|
|
)
|
|
|
|
monkeypatch.setattr("routers.ai.sam_registry.predict_points", fake_predict_points)
|
|
|
|
response = client.post("/api/ai/predict", json={
|
|
"image_id": frame["id"],
|
|
"prompt_type": "point",
|
|
"prompt_data": {"points": [[0.5, 0.5], [0.52, 0.52]], "labels": [1, 0]},
|
|
"options": {
|
|
"crop_to_prompt": True,
|
|
"crop_margin": 0.1,
|
|
"auto_filter_background": True,
|
|
"min_score": 0.05,
|
|
},
|
|
})
|
|
|
|
assert response.status_code == 200
|
|
assert calls["shape"][0] < 100
|
|
assert calls["shape"][1] < 200
|
|
assert calls["labels"] == [1, 0]
|
|
assert response.json()["scores"] == [0.9]
|
|
polygon = response.json()["polygons"][0]
|
|
assert all(0.0 <= coord <= 1.0 for point in polygon for coord in point)
|
|
|
|
|
|
def test_predict_box_and_semantic_fallback(client, monkeypatch):
|
|
_, frame, _ = _create_project_and_frame(client)
|
|
monkeypatch.setattr("routers.ai._load_frame_image", lambda frame: np.zeros((10, 10, 3), dtype=np.uint8))
|
|
monkeypatch.setattr("routers.ai.sam_registry.predict_box", lambda model, image, box: (
|
|
[[[0.2, 0.2], [0.8, 0.2], [0.8, 0.8]]],
|
|
[0.8],
|
|
))
|
|
monkeypatch.setattr("routers.ai.sam_registry.predict_semantic", lambda model, image, text: (
|
|
[[[0.0, 0.0], [1.0, 0.0], [1.0, 1.0]]],
|
|
[0.5],
|
|
))
|
|
|
|
box_response = client.post("/api/ai/predict", json={
|
|
"image_id": frame["id"],
|
|
"prompt_type": "box",
|
|
"prompt_data": [0.2, 0.2, 0.8, 0.8],
|
|
})
|
|
semantic_response = client.post("/api/ai/predict", json={
|
|
"image_id": frame["id"],
|
|
"prompt_type": "semantic",
|
|
"prompt_data": "胆囊",
|
|
})
|
|
|
|
assert box_response.status_code == 200
|
|
assert box_response.json()["scores"] == [0.8]
|
|
assert semantic_response.status_code == 200
|
|
assert semantic_response.json()["scores"] == [0.5]
|
|
|
|
|
|
def test_model_status_reports_runtime(client, monkeypatch):
|
|
monkeypatch.setattr("routers.ai.sam_registry.runtime_status", lambda selected_model=None: {
|
|
"selected_model": selected_model or "sam2",
|
|
"gpu": {
|
|
"available": False,
|
|
"device": "cpu",
|
|
"name": None,
|
|
"torch_available": True,
|
|
"torch_version": "2.x",
|
|
"cuda_version": None,
|
|
},
|
|
"models": [
|
|
{
|
|
"id": "sam2",
|
|
"label": "SAM 2",
|
|
"available": True,
|
|
"loaded": False,
|
|
"device": "cpu",
|
|
"supports": ["point", "box", "auto"],
|
|
"message": "ready",
|
|
"package_available": True,
|
|
"checkpoint_exists": True,
|
|
"checkpoint_path": "model.pt",
|
|
"python_ok": True,
|
|
"torch_ok": True,
|
|
"cuda_required": False,
|
|
},
|
|
{
|
|
"id": "sam3",
|
|
"label": "SAM 3",
|
|
"available": False,
|
|
"loaded": False,
|
|
"device": "unavailable",
|
|
"supports": ["semantic"],
|
|
"message": "missing Python 3.12+ runtime",
|
|
"package_available": False,
|
|
"checkpoint_exists": False,
|
|
"checkpoint_path": None,
|
|
"python_ok": False,
|
|
"torch_ok": True,
|
|
"cuda_required": True,
|
|
},
|
|
],
|
|
})
|
|
|
|
response = client.get("/api/ai/models/status?selected_model=sam3")
|
|
|
|
assert response.status_code == 200
|
|
body = response.json()
|
|
assert body["selected_model"] == "sam3"
|
|
assert body["models"][1]["id"] == "sam3"
|
|
assert body["models"][1]["available"] is False
|
|
|
|
|
|
def test_predict_validation_errors(client, monkeypatch):
|
|
project, _, _ = _create_project_and_frame(client)
|
|
|
|
assert client.post("/api/ai/predict", json={
|
|
"image_id": 999,
|
|
"prompt_type": "point",
|
|
"prompt_data": [[0.5, 0.5]],
|
|
}).status_code == 404
|
|
|
|
frame = client.post(f"/api/projects/{project['id']}/frames", json={
|
|
"project_id": project["id"],
|
|
"frame_index": 1,
|
|
"image_url": "frames/1.jpg",
|
|
}).json()
|
|
monkeypatch.setattr("routers.ai._load_frame_image", lambda frame: np.zeros((10, 10, 3), dtype=np.uint8))
|
|
assert client.post("/api/ai/predict", json={
|
|
"image_id": frame["id"],
|
|
"prompt_type": "box",
|
|
"prompt_data": [0.1, 0.2],
|
|
}).status_code == 400
|
|
|
|
|
|
def test_save_annotation_validates_project_and_frame(client):
|
|
project, frame, template = _create_project_and_frame(client)
|
|
|
|
saved = client.post("/api/ai/annotate", json={
|
|
"project_id": project["id"],
|
|
"frame_id": frame["id"],
|
|
"template_id": template["id"],
|
|
"mask_data": {"polygons": [[[0.1, 0.1], [0.9, 0.1], [0.9, 0.9]]]},
|
|
"points": [[0.5, 0.5]],
|
|
"bbox": [0.1, 0.1, 0.8, 0.8],
|
|
})
|
|
assert saved.status_code == 201
|
|
assert saved.json()["project_id"] == project["id"]
|
|
|
|
listing = client.get(f"/api/ai/annotations?project_id={project['id']}")
|
|
assert listing.status_code == 200
|
|
assert listing.json()[0]["id"] == saved.json()["id"]
|
|
|
|
frame_listing = client.get(f"/api/ai/annotations?project_id={project['id']}&frame_id={frame['id']}")
|
|
assert frame_listing.status_code == 200
|
|
assert len(frame_listing.json()) == 1
|
|
|
|
missing_project = client.post("/api/ai/annotate", json={"project_id": 999})
|
|
assert missing_project.status_code == 404
|
|
|
|
missing_frame = client.post("/api/ai/annotate", json={
|
|
"project_id": project["id"],
|
|
"frame_id": 999,
|
|
})
|
|
assert missing_frame.status_code == 404
|
|
|
|
missing_project_list = client.get("/api/ai/annotations?project_id=999")
|
|
assert missing_project_list.status_code == 404
|
|
|
|
|
|
def test_update_and_delete_annotation(client):
|
|
project, frame, template = _create_project_and_frame(client)
|
|
saved = client.post("/api/ai/annotate", json={
|
|
"project_id": project["id"],
|
|
"frame_id": frame["id"],
|
|
"template_id": template["id"],
|
|
"mask_data": {
|
|
"polygons": [[[0.1, 0.1], [0.9, 0.1], [0.9, 0.9]]],
|
|
"label": "AI Mask",
|
|
"color": "#06b6d4",
|
|
},
|
|
"points": [[0.5, 0.5]],
|
|
"bbox": [0.1, 0.1, 0.8, 0.8],
|
|
}).json()
|
|
|
|
updated = client.patch(f"/api/ai/annotations/{saved['id']}", json={
|
|
"template_id": template["id"],
|
|
"mask_data": {
|
|
"polygons": [[[0.2, 0.2], [0.8, 0.2], [0.8, 0.8]]],
|
|
"label": "胆囊",
|
|
"color": "#ff0000",
|
|
"class": {"id": "c1", "name": "胆囊", "color": "#ff0000", "zIndex": 20},
|
|
},
|
|
"points": [[0.4, 0.4]],
|
|
"bbox": [0.2, 0.2, 0.6, 0.6],
|
|
})
|
|
|
|
assert updated.status_code == 200
|
|
body = updated.json()
|
|
assert body["mask_data"]["label"] == "胆囊"
|
|
assert body["mask_data"]["class"]["id"] == "c1"
|
|
assert body["points"] == [[0.4, 0.4]]
|
|
assert body["bbox"] == [0.2, 0.2, 0.6, 0.6]
|
|
|
|
listing = client.get(f"/api/ai/annotations?project_id={project['id']}")
|
|
assert listing.status_code == 200
|
|
assert listing.json()[0]["mask_data"]["class"]["name"] == "胆囊"
|
|
|
|
deleted = client.delete(f"/api/ai/annotations/{saved['id']}")
|
|
assert deleted.status_code == 204
|
|
|
|
empty_listing = client.get(f"/api/ai/annotations?project_id={project['id']}")
|
|
assert empty_listing.status_code == 200
|
|
assert empty_listing.json() == []
|
|
|
|
|
|
def test_update_and_delete_annotation_validation(client):
|
|
project, frame, template = _create_project_and_frame(client)
|
|
saved = client.post("/api/ai/annotate", json={
|
|
"project_id": project["id"],
|
|
"frame_id": frame["id"],
|
|
"template_id": template["id"],
|
|
}).json()
|
|
|
|
assert client.patch("/api/ai/annotations/999", json={"bbox": [0, 0, 1, 1]}).status_code == 404
|
|
assert client.delete("/api/ai/annotations/999").status_code == 404
|
|
assert client.patch(
|
|
f"/api/ai/annotations/{saved['id']}",
|
|
json={"template_id": 999},
|
|
).status_code == 404
|
|
|
|
|
|
def test_import_gt_mask_creates_annotations_with_seed_points(client):
|
|
project, frame, template = _create_project_and_frame(client)
|
|
mask = np.zeros((360, 640), dtype=np.uint8)
|
|
cv2.rectangle(mask, (100, 80), (260, 220), 255, thickness=-1)
|
|
ok, encoded = cv2.imencode(".png", mask)
|
|
assert ok
|
|
|
|
response = client.post(
|
|
"/api/ai/import-gt-mask",
|
|
data={
|
|
"project_id": str(project["id"]),
|
|
"frame_id": str(frame["id"]),
|
|
"template_id": str(template["id"]),
|
|
"label": "Imported GT",
|
|
"color": "#22c55e",
|
|
},
|
|
files={"file": ("mask.png", encoded.tobytes(), "image/png")},
|
|
)
|
|
|
|
assert response.status_code == 201
|
|
body = response.json()
|
|
assert len(body) == 1
|
|
assert body[0]["project_id"] == project["id"]
|
|
assert body[0]["frame_id"] == frame["id"]
|
|
assert body[0]["template_id"] == template["id"]
|
|
assert body[0]["mask_data"]["label"] == "Imported GT"
|
|
assert body[0]["mask_data"]["source"] == "gt_mask"
|
|
assert body[0]["mask_data"]["gt_label_value"] == 255
|
|
assert len(body[0]["mask_data"]["polygons"][0]) >= 3
|
|
assert len(body[0]["points"]) == 1
|
|
assert 0.0 <= body[0]["points"][0][0] <= 1.0
|
|
assert 0.0 <= body[0]["points"][0][1] <= 1.0
|
|
|
|
|
|
def test_import_gt_mask_splits_label_values(client):
|
|
project, frame, _ = _create_project_and_frame(client)
|
|
mask = np.zeros((360, 640), dtype=np.uint8)
|
|
cv2.rectangle(mask, (20, 20), (120, 120), 1, thickness=-1)
|
|
cv2.rectangle(mask, (220, 80), (320, 180), 2, thickness=-1)
|
|
ok, encoded = cv2.imencode(".png", mask)
|
|
assert ok
|
|
|
|
response = client.post(
|
|
"/api/ai/import-gt-mask",
|
|
data={
|
|
"project_id": str(project["id"]),
|
|
"frame_id": str(frame["id"]),
|
|
"label": "GT Class",
|
|
},
|
|
files={"file": ("labels.png", encoded.tobytes(), "image/png")},
|
|
)
|
|
|
|
assert response.status_code == 201
|
|
body = sorted(response.json(), key=lambda item: item["mask_data"]["gt_label_value"])
|
|
assert [item["mask_data"]["gt_label_value"] for item in body] == [1, 2]
|
|
assert [item["mask_data"]["label"] for item in body] == ["GT Class 1", "GT Class 2"]
|
|
assert all(len(item["points"]) == 1 for item in body)
|