@camstack/addon-pipeline 1.2.332 → 1.2.334

package/dist/stream-broker/index.js

@@ -3,12 +3,12 @@ Object.defineProperties(exports, {
 	[Symbol.toStringTag]: { value: "Module" }
 });
 const require_chunk = require("../chunk-emK7D4bc.js");
-const require_dist = require("../dist-DJV7Z6Bb.js");
-const require_node = require("../node-CkmucLYG.js");
+const require_dist = require("../dist-C7v30DMP.js");
+const require_node = require("../node-DWdekhW3.js");
 const require_restream_intent = require("../restream-intent-7uDNqMKs.js");
 const require_remote_restream = require("../remote-restream-CO36Sr30.js");
 const require_retire_root_keys = require("../retire-root-keys-Csb8z2f3.js");
-const require_segment_demux_js = require("../segment-demux-js-mK9IBGLf.js");
+const require_segment_demux_js = require("../segment-demux-js-BqPTLFZG.js");
 let node_os = require("node:os");
 node_os = require_chunk.__toESM(node_os);
 let node_crypto = require("node:crypto");
@@ -24391,6 +24391,11 @@ var AdaptiveSession = class AdaptiveSession {
 	/** Throttle: minimum interval between PLI-triggered keyframe re-sends (ms). */
 	static PLI_RESEND_COOLDOWN_MS = 500;
 	lastPliResendAt = 0;
+	/** Minimum interval between two "PLI refused while recorded owns the
+	*  track" lines; the refusals in between are counted into the next one. */
+	static RECORDED_PLI_LOG_INTERVAL_MS = 1e4;
+	lastRecordedPliLogAt = 0;
+	recordedPliRefusedSinceLog = 0;
 	/**
 	* Per-session H.265 RTP repacketizer (lazy-init). The H.265 path
 	* forwards source RTP from the broker through this repacketizer so
@@ -25560,10 +25565,31 @@ var AdaptiveSession = class AdaptiveSession {
 			sessionId: this.sessionId
 		} });
 	}
+	logRecordedPliRefusal(now) {
+		this.recordedPliRefusedSinceLog += 1;
+		if (now - this.lastRecordedPliLogAt < AdaptiveSession.RECORDED_PLI_LOG_INTERVAL_MS) return;
+		this.lastRecordedPliLogAt = now;
+		const refused = this.recordedPliRefusedSinceLog;
+		this.recordedPliRefusedSinceLog = 0;
+		this.logger.info("PLI during recorded playback — live pre-buffer NOT replayed (recorded owns the track)", {
+			tags: { deviceId: this.deviceId },
+			meta: {
+				phase: "session",
+				sessionId: this.sessionId,
+				deviceId: this.deviceId,
+				refused,
+				windowMs: AdaptiveSession.RECORDED_PLI_LOG_INTERVAL_MS
+			}
+		});
+	}
 	handlePli() {
 		const now = Date.now();
 		if (now - this.lastPliResendAt < AdaptiveSession.PLI_RESEND_COOLDOWN_MS) return;
 		this.lastPliResendAt = now;
+		if (this.liveFeedGated) {
+			this.logRecordedPliRefusal(now);
+			return;
+		}
 		if (this.activeRtpReplay?.active) {
 			this.logger.debug("PLI received — paced pre-buffer replay already in flight", { meta: {
 				phase: "session",
@@ -26170,6 +26196,10 @@ var AdaptiveSession = class AdaptiveSession {
 	/** When gated, incoming live broker frames are dropped so recorded playback owns the track. */
 	setLiveFeedGate(gated) {
 		this.liveFeedGated = gated;
+		if (gated) {
+			this.activeRtpReplay?.abort();
+			this.activeRtpReplay = null;
+		}
 		this.rtpRebasePending = true;
 		this.recordedAudioRebasePending = true;
 	}

package/dist/stream-broker/index.mjs

@@ -1,10 +1,10 @@
 import { r as __toESM } from "../chunk-DnnnRqeS.mjs";
-import { $ as cameraStreamsCapability, $n as union, $t as resolveEgressDecodeHwAccel, At as logChannelsCapability, C as EncodeProfileSchema, Cn as errMsg, Ct as gifFromMp4Args, Dn as CamProfileSchema, E as G711_SCALE_CORRECTION_DB, En as CAM_PROFILE_ORDER, F as RATE_CONTROL_TIGHT, Fn as makeProfileBrokerId, Ft as memoizePinnedFfmpeg, Gn as discriminatedUnion, H as RingBuffer, Hn as _enum, In as makeSourceBrokerId, Jn as number, K as addonWidgetsSourceCapability, Kt as rankStreams, Ln as nodePin, Mn as expandAudioChunkToF32le, Mt as maskUrlCredentials, P as RATE_CONTROL_RELAXED, Pn as isEvent, Qn as string, Sn as DeviceType, St as gifDeriveSpeed, Tn as BaseAddon, Un as array, Wn as boolean, Yn as object, Z as batteryCapability, Zn as record, _n as buildFfmpegArgs$1, at as declareLogChannel, bn as isSoftwareDecode, bt as ffmpegRate, c as CLIP_STILL_REASON_CODE_HEADER, er as EventCategory, ft as egressTranscodeSharingKey, it as createLogChannelsProvider, kn as createEvent, kt as loadContributionCapability, ln as streamBrokerCapability, mn as AUDIO_PRESETS, o as BatteryStatusSchema, pn as webrtcSessionCapability, pt as egressTransportFromRequest, qn as literal, rt as createHwAccelCache, s as CLIP_BROKER_PACED_RATES, st as deriveBatteryPresence, tt as clipExportFailureOf, wn as BOOT_RECOVERY_BACKOFF_MS, xn as DeviceFeature, yn as invocationFromEncodeProfile, zn as parseProfileBrokerId } from "../dist-BsT15i7i.mjs";
-import { i as NO_COST_CLAIM, n as Fmp4FragmentChild, o as probeFfmpegBinary, r as Fmp4FragmentPlane, t as ChildCostRegistry } from "../node-Crsi_8ea.mjs";
+import { $ as cameraStreamsCapability, $n as union, $t as resolveEgressDecodeHwAccel, At as logChannelsCapability, C as EncodeProfileSchema, Cn as errMsg, Ct as gifFromMp4Args, Dn as CamProfileSchema, E as G711_SCALE_CORRECTION_DB, En as CAM_PROFILE_ORDER, F as RATE_CONTROL_TIGHT, Fn as makeProfileBrokerId, Ft as memoizePinnedFfmpeg, Gn as discriminatedUnion, H as RingBuffer, Hn as _enum, In as makeSourceBrokerId, Jn as number, K as addonWidgetsSourceCapability, Kt as rankStreams, Ln as nodePin, Mn as expandAudioChunkToF32le, Mt as maskUrlCredentials, P as RATE_CONTROL_RELAXED, Pn as isEvent, Qn as string, Sn as DeviceType, St as gifDeriveSpeed, Tn as BaseAddon, Un as array, Wn as boolean, Yn as object, Z as batteryCapability, Zn as record, _n as buildFfmpegArgs$1, at as declareLogChannel, bn as isSoftwareDecode, bt as ffmpegRate, c as CLIP_STILL_REASON_CODE_HEADER, er as EventCategory, ft as egressTranscodeSharingKey, it as createLogChannelsProvider, kn as createEvent, kt as loadContributionCapability, ln as streamBrokerCapability, mn as AUDIO_PRESETS, o as BatteryStatusSchema, pn as webrtcSessionCapability, pt as egressTransportFromRequest, qn as literal, rt as createHwAccelCache, s as CLIP_BROKER_PACED_RATES, st as deriveBatteryPresence, tt as clipExportFailureOf, wn as BOOT_RECOVERY_BACKOFF_MS, xn as DeviceFeature, yn as invocationFromEncodeProfile, zn as parseProfileBrokerId } from "../dist-Dn3knu_K.mjs";
+import { i as NO_COST_CLAIM, n as Fmp4FragmentChild, o as probeFfmpegBinary, r as Fmp4FragmentPlane, t as ChildCostRegistry } from "../node-_BruMF4L.mjs";
 import { a as parseRestreamPath, d as PREBUFFER_MAX_MS, f as capPreRollRequestMs, i as isLiveEdgeOnlyDial, m as resolvePrebuffer, n as NO_DECLARATION, p as prebufferRetentionMs, s as sdpVariantFor, u as withSentryIntent } from "../restream-intent-DyGmLAU9.mjs";
 import { n as profileForStreamId } from "../remote-restream-BeHi78PZ.mjs";
 import { n as RowMapStore, t as retireRootKeys } from "../retire-root-keys-DZvlso2l.mjs";
-import { a as require_mp4box_all, c as parseNalLengthSize, i as runFfmpegOnce, n as demuxSegmentPreferJs, o as avccSampleToAnnexB, r as FfmpegFallbackRefusedError, s as buildKeyframeAnnexB, t as demuxSegmentFullJs } from "../segment-demux-js-B_tboBcJ.mjs";
+import { a as require_mp4box_all, c as parseNalLengthSize, i as runFfmpegOnce, n as demuxSegmentPreferJs, o as avccSampleToAnnexB, r as FfmpegFallbackRefusedError, s as buildKeyframeAnnexB, t as demuxSegmentFullJs } from "../segment-demux-js-B_-zBECg.mjs";
 import { createRequire } from "node:module";
 import * as os from "node:os";
 import { networkInterfaces, tmpdir } from "node:os";
@@ -24387,6 +24387,11 @@ var AdaptiveSession = class AdaptiveSession {
 	/** Throttle: minimum interval between PLI-triggered keyframe re-sends (ms). */
 	static PLI_RESEND_COOLDOWN_MS = 500;
 	lastPliResendAt = 0;
+	/** Minimum interval between two "PLI refused while recorded owns the
+	*  track" lines; the refusals in between are counted into the next one. */
+	static RECORDED_PLI_LOG_INTERVAL_MS = 1e4;
+	lastRecordedPliLogAt = 0;
+	recordedPliRefusedSinceLog = 0;
 	/**
 	* Per-session H.265 RTP repacketizer (lazy-init). The H.265 path
 	* forwards source RTP from the broker through this repacketizer so
@@ -25556,10 +25561,31 @@ var AdaptiveSession = class AdaptiveSession {
 			sessionId: this.sessionId
 		} });
 	}
+	logRecordedPliRefusal(now) {
+		this.recordedPliRefusedSinceLog += 1;
+		if (now - this.lastRecordedPliLogAt < AdaptiveSession.RECORDED_PLI_LOG_INTERVAL_MS) return;
+		this.lastRecordedPliLogAt = now;
+		const refused = this.recordedPliRefusedSinceLog;
+		this.recordedPliRefusedSinceLog = 0;
+		this.logger.info("PLI during recorded playback — live pre-buffer NOT replayed (recorded owns the track)", {
+			tags: { deviceId: this.deviceId },
+			meta: {
+				phase: "session",
+				sessionId: this.sessionId,
+				deviceId: this.deviceId,
+				refused,
+				windowMs: AdaptiveSession.RECORDED_PLI_LOG_INTERVAL_MS
+			}
+		});
+	}
 	handlePli() {
 		const now = Date.now();
 		if (now - this.lastPliResendAt < AdaptiveSession.PLI_RESEND_COOLDOWN_MS) return;
 		this.lastPliResendAt = now;
+		if (this.liveFeedGated) {
+			this.logRecordedPliRefusal(now);
+			return;
+		}
 		if (this.activeRtpReplay?.active) {
 			this.logger.debug("PLI received — paced pre-buffer replay already in flight", { meta: {
 				phase: "session",
@@ -26166,6 +26192,10 @@ var AdaptiveSession = class AdaptiveSession {
 	/** When gated, incoming live broker frames are dropped so recorded playback owns the track. */
 	setLiveFeedGate(gated) {
 		this.liveFeedGated = gated;
+		if (gated) {
+			this.activeRtpReplay?.abort();
+			this.activeRtpReplay = null;
+		}
 		this.rtpRebasePending = true;
 		this.recordedAudioRebasePending = true;
 	}

package/dist/stream-broker/remoteEntry.js

@@ -30,7 +30,7 @@ async function d(e) {
 	}
 }
 async function f() {
-	return l ||= d(() => import("./_virtual_mf-localSharedImportMap___mfe_internal__addon_stream_broker_widgets-CC0JSUlW.mjs")).catch((e) => {
+	return l ||= d(() => import("./_virtual_mf-localSharedImportMap___mfe_internal__addon_stream_broker_widgets-BP8FFU2i.mjs")).catch((e) => {
 		throw l = void 0, e;
 	}), l;
 }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@camstack/addon-pipeline",
-  "version": "1.2.332",
+  "version": "1.2.334",
   "description": "Pipeline bundle — runner, detection, motion, audio + stream broker. Multi-entry npm package shipping pipeline addons under a single bundle.",
   "keywords": [
     "camstack",

package/python/postprocessors/test_yolo_seg_root.py

@@ -0,0 +1,109 @@
+"""D737 — a YOLO26-seg ROOT decodes its boxes exactly like the plain detector
+and keeps one mask per kept detection.
+
+Run like the other postprocessor tests:
+``PYTHONPATH=postprocessors:. python -m pytest postprocessors/`` from
+``packages/addon-pipeline/python``.
+"""
+import base64
+
+import numpy as np
+
+from postprocessors.yolo import postprocess_yolo
+from postprocessors.yolo_seg import postprocess_yolo_seg
+
+LABELS = ["person", "car", "dog"]
+CONFIG = {"labels": LABELS, "numClasses": 3, "confidence": 0.3, "inputSize": 64}
+# A 1280x720 frame letterboxed into 64: scale 0.05, vertical pad 14.
+ORIG_W, ORIG_H, SCALE, PAD = 1280, 720, 0.05, (0, 14)
+PROTO = 16  # P x P prototype grid at the 64 input (4 px per cell)
+
+
+def _protos(rng: np.random.Generator) -> np.ndarray:
+    return rng.normal(size=(1, 32, PROTO, PROTO)).astype(np.float32)
+
+
+def _post_nms_rows(rng: np.random.Generator) -> np.ndarray:
+    """[1, 6, 38]: two overlapping persons (the end-to-end head keeps both),
+    a car, one row under the floor and two padding rows (class -1)."""
+    boxes = [
+        [8, 20, 24, 44, 0.91, 0],
+        [10, 22, 26, 46, 0.80, 0],
+        [30, 30, 56, 48, 0.66, 1],
+        [40, 18, 50, 28, 0.10, 2],
+        [0, 0, 0, 0, 0.0, -1],
+        [0, 0, 0, 0, 0.0, -1],
+    ]
+    coeffs = rng.normal(size=(len(boxes), 32))
+    return np.concatenate([np.array(boxes, dtype=np.float32), coeffs], axis=1)[None].astype(np.float32)
+
+
+def _raw_head(rng: np.random.Generator) -> np.ndarray:
+    """[1, 4 + 3 + 32, 5] with LOGIT scores (YOLO26 raw), two overlapping."""
+    cxcywh = np.array(
+        [[16, 32, 16, 24], [18, 34, 16, 24], [43, 39, 26, 18], [45, 23, 10, 10], [5, 5, 4, 4]],
+        dtype=np.float32,
+    ).T
+    logits = np.array(
+        [[3.0, 2.0, -4, -4, -6], [-4, -4, 1.5, -4, -6], [-4, -4, -4, -2.0, -6]], dtype=np.float32
+    )
+    coeffs = rng.normal(size=(32, 5)).astype(np.float32)
+    return np.concatenate([cxcywh, logits, coeffs], axis=0)[None]
+
+
+def _boxes(out: dict) -> list[tuple]:
+    return [(d["class"], d["score"], tuple(d["bbox"])) for d in out["detections"]]
+
+
+def test_post_nms_head_boxes_equal_the_plain_detectors():
+    rng = np.random.default_rng(1)
+    rows = _post_nms_rows(rng)
+    plain = postprocess_yolo({"output0": rows[..., :6].copy()}, dict(CONFIG), ORIG_W, ORIG_H, SCALE, PAD)
+    seg = postprocess_yolo_seg(
+        {"output0": rows, "output1": _protos(rng)}, dict(CONFIG), ORIG_W, ORIG_H, SCALE, PAD
+    )
+    assert _boxes(seg) == _boxes(plain)
+    # The plain detector does not NMS an end-to-end head; neither does the seg root.
+    assert len(seg["detections"]) == 3
+
+
+def test_raw_head_boxes_equal_the_plain_detectors():
+    rng = np.random.default_rng(2)
+    head = _raw_head(rng)
+    plain = postprocess_yolo({"output0": head[:, :7, :].copy()}, dict(CONFIG), ORIG_W, ORIG_H, SCALE, PAD)
+    seg = postprocess_yolo_seg(
+        {"output0": head.copy(), "output1": _protos(rng)}, dict(CONFIG), ORIG_W, ORIG_H, SCALE, PAD
+    )
+    assert _boxes(seg) == _boxes(plain)
+    assert len(seg["detections"]) == 2  # the overlap is suppressed, the floor drops two
+
+
+def test_same_nms_iou_knob_as_the_plain_detector():
+    rng = np.random.default_rng(3)
+    head = _raw_head(rng)
+    loose = dict(CONFIG, nmsIouThreshold=0.95)
+    plain = postprocess_yolo({"output0": head[:, :7, :].copy()}, loose, ORIG_W, ORIG_H, SCALE, PAD)
+    seg = postprocess_yolo_seg(
+        {"output0": head.copy(), "output1": _protos(rng)}, loose, ORIG_W, ORIG_H, SCALE, PAD
+    )
+    assert _boxes(seg) == _boxes(plain)
+    assert len(seg["detections"]) == 3
+
+
+def test_every_kept_detection_carries_its_own_mask_over_its_box():
+    rng = np.random.default_rng(4)
+    rows = _post_nms_rows(rng)
+    protos = _protos(rng)
+    out = postprocess_yolo_seg({"output0": rows, "output1": protos}, dict(CONFIG), ORIG_W, ORIG_H, SCALE, PAD)
+    first = out["detections"][0]
+    # Box [8,20,24,44] in the 64 input → proto cells [2,5,6,11] → 4 x 6.
+    assert (first["maskWidth"], first["maskHeight"]) == (4, 6)
+    raw = np.frombuffer(base64.b64decode(first["mask"]), dtype=np.uint8)
+    assert raw.size == 4 * 6
+    assert set(np.unique(raw)) <= {0, 255}
+    # The mask IS sigmoid(coeffs @ protos) > 0.5 over that window, for this row.
+    logits = rows[0, 0, 6:] @ protos[0][:, 5:11, 2:6].reshape(32, -1)
+    expected = ((1 / (1 + np.exp(-logits))) > 0.5).astype(np.uint8) * 255
+    assert np.array_equal(raw, expected)
+    # Different rows → different coefficient vectors → masks are per instance.
+    assert out["detections"][1]["mask"] != first["mask"]

package/python/postprocessors/yolo.py

@@ -180,12 +180,28 @@ def _postprocess_yolo11_nms(
 
     Coordinates are in input pixel space (0..inputSize). Already post-NMS.
     """
+    rows = decode_yolo11_nms_rows(output, config, orig_w, orig_h, scale, pad)
+    return {"kind": "detections", "detections": [d for _, d in rows]}
+
+
+def decode_yolo11_nms_rows(
+    output: np.ndarray,
+    config: dict,
+    orig_w: int,
+    orig_h: int,
+    scale: float,
+    pad: tuple[int, int],
+) -> list[tuple[int, dict]]:
+    """The post-NMS box decode, each kept detection paired with its ROW in
+    ``output``. Only the first 6 columns are read, so a seg head's
+    ``[N, 6 + 32]`` rows decode exactly like the plain detector's ``[N, 6]``
+    (D737) and the row index finds the detection's mask coefficients."""
     conf_threshold = float(config.get("confidence", 0))
     labels = config.get("labels", COCO_80)
 
-    detections = []
+    detections: list[tuple[int, dict]] = []
     for i in range(output.shape[0]):
-        x1, y1, x2, y2, score, class_id = output[i]
+        x1, y1, x2, y2, score, class_id = output[i][:6]
         score = float(score)
         if score < conf_threshold:
             continue
@@ -200,13 +216,13 @@ def _postprocess_yolo11_nms(
         oy2 = max(0, min(orig_h, (float(y2) - pad[1]) / scale))
 
         label = labels[class_id] if class_id < len(labels) else str(class_id)
-        detections.append({
+        detections.append((i, {
             "class": label,
             "score": round(score, 4),
             "bbox": [round(ox1, 1), round(oy1, 1), round(ox2, 1), round(oy2, 1)],
-        })
+        }))
 
-    return {"kind": "detections", "detections": detections}
+    return detections
 
 
 def postprocess_yolo(
@@ -223,12 +239,8 @@ def postprocess_yolo(
     1. Raw tensor (ONNX/OpenVINO): [4+numClasses, numBoxes] — needs NMS
     2. CoreML built-in NMS: 'confidence' + 'coordinates' tensors — already post-NMS
     """
-    conf_threshold = float(config.get("confidence", 0))
     labels = config.get("labels", COCO_80)
     num_classes = config.get("numClasses", len(labels))
-    # NMS IoU from the step config (operator slider); the 0.45 literal stays as
-    # the fallback so an unset value is byte-identical to today (#4).
-    nms_iou = float(config.get("nmsIouThreshold", 0.45))
 
     yolox = yolox_head_to_v8(predictions, int(config.get("inputSize", 640)))
     if yolox is not None:
@@ -272,6 +284,31 @@ def postprocess_yolo(
         else:
             output = output.T
 
+    rows = decode_raw_yolo_columns(output, config, orig_w, orig_h, scale, pad)
+    return {"kind": "detections", "detections": [d for _, d in rows]}
+
+
+def decode_raw_yolo_columns(
+    output: np.ndarray,
+    config: dict,
+    orig_w: int,
+    orig_h: int,
+    scale: float,
+    pad: tuple[int, int],
+) -> list[tuple[int, dict]]:
+    """The raw-head box decode (``[4 + numClasses (+ extra), numBoxes]``, one
+    COLUMN per candidate): sigmoid when the scores are logits, confidence
+    floor, NMS. Each kept detection is paired with its column in ``output``.
+    Rows past ``4 + numClasses`` are never read, so a seg head's mask
+    coefficients ride along untouched and decode exactly like the plain
+    detector (D737). May apply the sigmoid IN PLACE on the class rows."""
+    conf_threshold = float(config.get("confidence", 0))
+    labels = config.get("labels", COCO_80)
+    num_classes = config.get("numClasses", len(labels))
+    # NMS IoU from the step config (operator slider); the 0.45 literal stays as
+    # the fallback so an unset value is byte-identical to today (#4).
+    nms_iou = float(config.get("nmsIouThreshold", 0.45))
+
     num_boxes = output.shape[1]
 
     # Detect if scores are raw logits (YOLO11/v26) vs sigmoid-activated (YOLOv8/v9).
@@ -290,7 +327,7 @@ def postprocess_yolo(
     best_scores = class_scores[best_classes, np.arange(num_boxes)]  # (N,)
     mask = best_scores >= conf_threshold
     if not np.any(mask):
-        return {"kind": "detections", "detections": []}
+        return []
 
     # Gather only above-threshold boxes
     idxs = np.nonzero(mask)[0]
@@ -315,11 +352,13 @@ def postprocess_yolo(
             "score": round(float(sc[k]), 4),
             "bbox": [round(float(x1[k]), 1), round(float(y1[k]), 1), round(float(x2[k]), 1), round(float(y2[k]), 1)],
             "_xyxy": (float(x1[k]), float(y1[k]), float(x2[k]), float(y2[k])),
+            "_col": int(idxs[k]),
         })
 
     kept = _nms(candidates, nms_iou)
-    # Remove internal _xyxy field
+    # Remove the internal fields; the column travels beside the detection.
+    out: list[tuple[int, dict]] = []
     for d in kept:
         del d["_xyxy"]
-
-    return {"kind": "detections", "detections": kept}
+        out.append((d.pop("_col"), d))
+    return out