File size: 9,305 Bytes
7e25f7a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
"""
Correlation Engine service — correlates evidence across all providers
into a deterministic relationship graph.

Takes a UnifiedFaceReport (the output of any analysis job) and:
  1. Extracts nodes: faces, objects, metadata, images, embeddings.
  2. Runs deterministic matchers to find relationships.
  3. Builds a CorrelationGraph with typed edges.

No AI — only deterministic matching via cores.correlation.matchers.

Usage:
  POST /analysis/correlate with an image → runs full pipeline → correlates
  POST /analysis/correlate with a pre-computed report → correlates only
"""

from __future__ import annotations

import time
from typing import Optional

import numpy as np

from cores.correlation import (
    CorrelationGraphBuilder,
    Node,
    EdgeType,
    match_faces,
    match_objects,
    match_locations,
    match_cameras,
    match_hashes,
    match_embeddings,
    match_metadata,
    match_timestamps,
)
from cores.vision import sha256_bytes, phash
from models.reports import UnifiedFaceReport
from pipeline import InputValidator, ImagePreprocessor, ImageHasher
from utils.logging import execution_context, new_execution_id


class CorrelationEngineService:
    """Correlates evidence across providers into a relationship graph."""

    def __init__(
        self,
        validator: InputValidator,
        preprocessor: ImagePreprocessor,
        hasher: ImageHasher,
    ) -> None:
        self._validator = validator
        self._preprocessor = preprocessor
        self._hasher = hasher

    async def correlate(self, report: dict) -> dict:
        """Correlate a pre-computed report.

        Args:
            report: a UnifiedFaceReport as a dict (or the report sub-object).

        Returns:
            {"success": True, "correlation_graph": {...}, "elapsed_ms": float}
        """
        eid = new_execution_id()
        with execution_context(execution_id=eid, provider_id="correlation_engine"):
            t0 = time.perf_counter()

            builder = CorrelationGraphBuilder()

            # --- Extract nodes ---

            # Face nodes (from detections)
            face_embeddings: list[tuple[str, np.ndarray]] = []
            for i, det in enumerate(report.get("detections", [])):
                node_id = f"face_{i}"
                builder.add_node(Node(
                    id=node_id,
                    node_type="face",
                    label=f"Face {i}",
                    properties={"box": det.get("box", {})},
                ))
                if det.get("embedding"):
                    face_embeddings.append((node_id, np.array(det["embedding"])))

            # Object nodes (from object_detections + object_intelligence)
            objects_for_matching: list[tuple[str, str, dict]] = []
            # From object_intelligence (richer)
            obj_intel = report.get("object_intelligence") or {}
            for i, obj in enumerate(obj_intel.get("objects", [])):
                node_id = f"object_{i}"
                builder.add_node(Node(
                    id=node_id,
                    node_type="object",
                    label=obj.get("class_label", "object"),
                    properties={"box": obj.get("box", {}), "confidence": obj.get("confidence", 0)},
                ))
                objects_for_matching.append((node_id, obj.get("class_label", ""), obj.get("box", {})))
            # Fallback: from object_detections (provider-level)
            if not objects_for_matching:
                for od in report.get("object_detections", []):
                    for i, obj in enumerate(od.get("objects", [])):
                        node_id = f"object_{od.get('provider', 'x')}_{i}"
                        builder.add_node(Node(
                            id=node_id,
                            node_type="object",
                            label=obj.get("label", "object"),
                            properties={"box": obj.get("box", {})},
                        ))
                        objects_for_matching.append((node_id, obj.get("label", ""), obj.get("box", {})))

            # Metadata nodes (from metadata_extractions + forensic_metadata)
            metadata_for_matching: list[tuple[str, dict]] = []
            for i, meta in enumerate(report.get("metadata_extractions", [])):
                node_id = f"metadata_{i}"
                builder.add_node(Node(
                    id=node_id,
                    node_type="metadata",
                    label=meta.get("format", "metadata"),
                    properties={"provider": meta.get("provider", "")},
                ))
                metadata_for_matching.append((node_id, meta.get("exif", {})))
            # Forensic metadata
            forensic_meta = report.get("forensic_metadata")
            if forensic_meta:
                node_id = "forensic_metadata"
                builder.add_node(Node(
                    id=node_id,
                    node_type="metadata",
                    label="forensic_metadata",
                    properties={"camera_make": forensic_meta.get("camera_make"),
                                "camera_model": forensic_meta.get("camera_model")},
                ))
                metadata_for_matching.append((node_id, forensic_meta.get("exif", {})))

            # Image nodes (hash)
            image_hashes: list[tuple[str, str, str]] = []
            image_hash = report.get("metadata", {}).get("image_hash")
            if image_hash:
                node_id = "image_0"
                builder.add_node(Node(
                    id=node_id,
                    node_type="image",
                    label="source_image",
                    properties={"sha256": image_hash},
                ))
                image_hashes.append((node_id, image_hash, ""))  # phash empty if not available

            # Embedding nodes
            image_embeddings: list[tuple[str, np.ndarray]] = []
            for i, emb in enumerate(report.get("embedding_results", [])):
                node_id = f"embedding_{i}"
                builder.add_node(Node(
                    id=node_id,
                    node_type="embedding",
                    label=emb.get("model", "embedding"),
                    properties={"dimensions": emb.get("dimensions", 0)},
                ))
                if emb.get("embedding"):
                    image_embeddings.append((node_id, np.array(emb["embedding"])))

            # Location nodes
            location_items: list[tuple[str, dict]] = []
            loc_est = report.get("location_estimate")
            if loc_est and loc_est.get("gps"):
                node_id = "location_0"
                builder.add_node(Node(
                    id=node_id,
                    node_type="location",
                    label="gps_location",
                    properties=loc_est["gps"],
                ))
                location_items.append((node_id, loc_est["gps"]))

            # --- Run matchers ---

            if face_embeddings:
                builder.add_matches(EdgeType.SAME_FACE, match_faces(face_embeddings))

            if objects_for_matching:
                builder.add_matches(EdgeType.SAME_OBJECT, match_objects(objects_for_matching))

            if location_items:
                builder.add_matches(EdgeType.SAME_LOCATION, match_locations(location_items))

            if image_hashes:
                builder.add_matches(EdgeType.SAME_HASH, match_hashes(image_hashes))

            if image_embeddings:
                builder.add_matches(EdgeType.SAME_EMBEDDING, match_embeddings(image_embeddings))

            if metadata_for_matching:
                builder.add_matches(EdgeType.SAME_METADATA, match_metadata(metadata_for_matching))
                # Camera matching
                cameras: list[tuple[str, str, str, str]] = []
                for item_id, exif in metadata_for_matching:
                    make = exif.get("Make", "")
                    model = exif.get("Model", "")
                    fp = ""  # fingerprint would come from forensic_metadata
                    cameras.append((item_id, make, model, fp))
                if forensic_meta and forensic_meta.get("camera_fingerprint"):
                    cameras.append(("forensic_metadata",
                                    forensic_meta.get("camera_make", ""),
                                    forensic_meta.get("camera_model", ""),
                                    forensic_meta["camera_fingerprint"]))
                builder.add_matches(EdgeType.SAME_CAMERA, match_cameras(cameras))

                # Timestamp matching
                timestamps: list[tuple[str, str]] = []
                for item_id, exif in metadata_for_matching:
                    ts = exif.get("DateTimeOriginal") or exif.get("DateTime")
                    if ts:
                        timestamps.append((item_id, str(ts)))
                if timestamps:
                    builder.add_matches(EdgeType.SAME_TIMESTAMP, match_timestamps(timestamps))

            elapsed = (time.perf_counter() - t0) * 1000.0
            graph = builder.build(elapsed_ms=elapsed)

            return {
                "success": True,
                "correlation_graph": graph.model_dump(),
                "elapsed_ms": round(elapsed, 3),
            }