Intel/delivery-package-verification

Delivery/Package Verification

Property	Value
Category	Object Detection (Package and Parcel Detection)
Base Model	YOLO26 (Ultralytics, default) or DETR (Facebook, alternative)
Source Framework	PyTorch (Ultralytics, Hugging Face Transformers)
Supported Precisions	FP32, FP16, INT8 (mixed-precision)
Inference Engine	OpenVINO
Hardware	CPU, GPU, NPU
Detected Class(es)	package (COCO backpack/handbag/suitcase, relabeled)

---

Overview

Delivery/Package Verification is a Metro Analytics use case that detects and counts delivery parcels, bags, and luggage items in camera feeds. It is built on YOLO26, a state-of-the-art real-time object detector trained on the COCO dataset, quantized to INT8 and filtered at runtime to the COCO classes that best match delivery parcels -- backpack, handbag, and suitcase -- which are all relabeled to a single package class in the output.

These COCO classes provide reliable coverage for typical delivery and package verification scenarios (for example a courier carrying a cardboard box) without requiring a custom-trained model. For label or text reading on packages, pair this with the ocr-text-recognition use case.

Typical Metro deployments include:

• Delivery Dock Monitoring -- verify parcels placed or removed at a loading area.
• Abandoned Luggage Detection -- flag unattended bags on platforms.
• Package Counting -- count parcels on a conveyor or at a drop-off zone.
• Theft Prevention -- alert when a package disappears from a monitored area.

Available variants: yolo26n, yolo26s, yolo26m, yolo26l. Smaller variants (yolo26n, yolo26s) are recommended for high-FPS edge deployment; larger variants improve recall for distant or partially occluded packages.

Model Selection

DETR (DEtection TRansformer) is available as an alternative to YOLO26. DETR is a transformer-based, NMS-free detector trained on COCO and released under the Apache-2.0 license. The export script embeds DETR normalization, box decoding, and the class remap into the OpenVINO IR, so the DETR model produces the same output layout and class ids as YOLO26. The same OpenVINO and DLStreamer samples run either model; select one with the MODEL constant at the top of each sample.

Model	Variants	Input Size	Output	License
YOLO26 (default)	yolo26n, yolo26s, yolo26m, yolo26l	640 x 640	[1, 300, 6]	AGPL-3.0
DETR	detr-resnet-50	800 x 800	[1, 100, 6]	Apache-2.0

Choose YOLO26 for maximum throughput on edge devices. Choose DETR when a permissively licensed transformer detector is preferred; it is heavier and runs at a lower frame rate.

---

Prerequisites

Target stack: Intel DLStreamer 2026.3.0 RC1 with the matching OpenVINO GenAI bundle 2026.5.0.0.dev20261030 (OpenVINO Runtime 2026.5.0.dev20261030). Use the native RC1 setup before running these samples; this is a development stack, not a stable release. See the device validation matrix for measured CPU, GPU, and NPU results.

• Python 3.11+
• Install OpenVINO (latest version)
• Install Intel DLStreamer (latest version)

Create and activate a Python virtual environment before running the scripts:

python3 -m venv .venv --system-site-packages
source .venv/bin/activate

Note: The --system-site-packages flag is required so the virtual environment can access the system-installed OpenVINO and DLStreamer Python packages.

---

Getting Started

Download and Quantize Model

Run the provided script to download, export to OpenVINO IR, and optionally quantize:

chmod +x export_and_quantize.sh
./export_and_quantize.sh

This exports the default yolo26n model in FP16 precision.

Optional: Select a Different Model, Variant, or Precision

./export_and_quantize.sh yolo26n FP32          # full-precision
./export_and_quantize.sh yolo26n INT8          # quantized
./export_and_quantize.sh yolo26s               # larger variant, default FP16
./export_and_quantize.sh detr-resnet-50        # DETR alternative, default FP16
./export_and_quantize.sh detr-resnet-50 INT8   # DETR quantized

The first argument selects the model: any YOLO26 variant (yolo26n, yolo26s, yolo26m, yolo26l) or DETR (detr-resnet-50). The script performs the following steps:

1. Installs dependencies (openvino, ultralytics for YOLO26 or transformers, torch for DETR; adds nncf for INT8).
2. Downloads a sample test image (test.jpg) and a sample test video (test_video.mp4).
3. Downloads the PyTorch weights and exports to OpenVINO IR.
4. (INT8 only) Quantizes the model using NNCF post-training quantization.

Output files (replace yolo26n with the selected model, for example detr-resnet-50):

• yolo26n_openvino_model/ -- FP32 or FP16 OpenVINO IR model directory.
• yolo26n_package_int8.xml / .bin -- INT8 quantized model (only when INT8 is selected).

Precision / Device Compatibility

Precision	CPU	GPU	NPU
FP32	Yes	Yes	No
FP16	Yes	Yes	Yes
INT8	Yes	Yes	Yes

OpenVINO Sample

The sample below runs YOLO26 inference on the sample video, filters detections to the delivery-package classes (COCO backpack, handbag, suitcase, all shown as package), annotates each frame, and writes the result to output_openvino.mp4 while printing the package count per frame. Frames are letterboxed (aspect-ratio-preserving resize with padding) before inference so the input matches how DLStreamer's gvadetect preprocesses. YOLO26 is end-to-end (NMS-free), so no manual non-maximum suppression is needed. Change the device string to run on CPU, GPU, or NPU.

The full sample is in openvino_sample.py; run it with python3 openvino_sample.py or copy the code below.

# SPDX-License-Identifier: MIT
# Copyright (C) Intel Corporation
#
# Delivery/Package Verification: OpenVINO Runtime sample.
# See README.md for prerequisites and inputs. Run: python3 openvino_sample.py

import cv2
import numpy as np
import openvino as ov

# COCO classes used as delivery-package proxies; all shown as "package".
PACKAGE_CLASS_IDS = {24, 26, 28}  # backpack, handbag, suitcase
PACKAGE_LABEL = "package"
BOX_COLOR = (0, 200, 0)
CONF_THRESHOLD = 0.25
# Select the model: "yolo26n" (default) or "detr-resnet-50".
MODEL = "yolo26n"
INPUT_SIZE = 800 if MODEL.startswith("detr") else 640
INPUT_VIDEO = "test_video.mp4"

core = ov.Core()
model = core.read_model(f"{MODEL}_openvino_model/{MODEL}.xml")

# Change device to "GPU" or "NPU" to run on integrated GPU or NPU.
compiled = core.compile_model(model, "CPU")


def letterbox(image, size=INPUT_SIZE):
    """Resize keeping aspect ratio and pad to a square (matches gvadetect)."""
    h, w = image.shape[:2]
    ratio = min(size / h, size / w)
    nw, nh = int(round(w * ratio)), int(round(h * ratio))
    resized = cv2.resize(image, (nw, nh))
    canvas = np.full((size, size, 3), 114, dtype=np.uint8)
    pad_x, pad_y = (size - nw) // 2, (size - nh) // 2
    canvas[pad_y:pad_y + nh, pad_x:pad_x + nw] = resized
    return canvas, ratio, pad_x, pad_y


cap = cv2.VideoCapture(INPUT_VIDEO)
fps = cap.get(cv2.CAP_PROP_FPS) or 30.0
width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
writer = cv2.VideoWriter(
    "output_openvino.mp4", cv2.VideoWriter_fourcc(*"mp4v"), fps, (width, height))

frame_idx = 0
while True:
    ok, frame = cap.read()
    if not ok:
        break
    frame_idx += 1

    padded, ratio, pad_x, pad_y = letterbox(frame, INPUT_SIZE)
    blob = cv2.cvtColor(padded, cv2.COLOR_BGR2RGB).astype(np.float32) / 255.0
    blob = blob.transpose(2, 0, 1)[np.newaxis, ...]  # NCHW

    # End-to-end output: [1, N, 6] = [x1, y1, x2, y2, confidence, class_id]
    # (N = 300 for YOLO26, 100 for DETR)
    output = compiled([blob])[compiled.output(0)][0]
    mask = (output[:, 4] >= CONF_THRESHOLD) & np.isin(
        output[:, 5].astype(int), list(PACKAGE_CLASS_IDS))
    dets = output[mask]

    for det in dets:
        # Undo the letterbox padding and scaling to map boxes back to the frame.
        x1 = int((det[0] - pad_x) / ratio)
        y1 = int((det[1] - pad_y) / ratio)
        x2 = int((det[2] - pad_x) / ratio)
        y2 = int((det[3] - pad_y) / ratio)
        conf = float(det[4])
        label = f"{PACKAGE_LABEL} {conf:.2f}"
        cv2.rectangle(frame, (x1, y1), (x2, y2), BOX_COLOR, 2)
        cv2.putText(frame, label, (x1, y1 - 5),
                    cv2.FONT_HERSHEY_SIMPLEX, 0.6, BOX_COLOR, 2)

    writer.write(frame)
    print(f"Frame {frame_idx}: Packages detected: {len(dets)}", flush=True)

cap.release()
writer.release()

Device targets:

• "CPU" -- default, works on all Intel platforms.
• "GPU" -- Intel integrated or discrete GPU.
• "NPU" -- Intel NPU (validate with benchmark_app -d NPU).

Expected Output

DLStreamer Sample

The pipeline below runs the FP16 YOLO26 detector on the sample video via gvadetect, renders only package bounding boxes using gvawatermark with displ-cfg=show-roi=package, saves the annotated result to output_dlstreamer.mp4, and prints the package count per frame.

Notes on running this sample:

• Use the FP16 IR (yolo26n_openvino_model/yolo26n.xml or detr-resnet-50_openvino_model/detr-resnet-50.xml) together with the coco_package_labels.txt label map produced by export_and_quantize.sh. It relabels the COCO backpack/handbag/suitcase classes to package, so gvadetect emits a single package class and gvawatermark renders a package label.

• Export PYTHONPATH so the DLStreamer Python module is importable:

  source /opt/intel/openvino_2026/setupvars.sh
  source /opt/intel/dlstreamer/scripts/setup_dls_env.sh
  export PYTHONPATH=/opt/intel/dlstreamer/python:\
  /opt/intel/dlstreamer/gstreamer/lib/python3/dist-packages:${PYTHONPATH:-}
  

The full sample is in dlstreamer_sample.py; run it with python3 dlstreamer_sample.py or copy the code below.

# SPDX-License-Identifier: MIT
# Copyright (C) Intel Corporation
#
# Delivery/Package Verification: DLStreamer pipeline sample.
# See README.md for prerequisites and inputs. Run: python3 dlstreamer_sample.py

import gi

gi.require_version("Gst", "1.0")
gi.require_version("GstAnalytics", "1.0")
from gi.repository import Gst, GLib, GstAnalytics

Gst.init([])

INPUT_VIDEO = "test_video.mp4"
PACKAGE_LABELS = {"package"}

# Select the model: "yolo26n" (default) or "detr-resnet-50".
MODEL = "yolo26n"

# For CPU: change device=GPU to device=CPU.
# For NPU: change device=GPU to device=NPU (batch-size=1, nireq=4 recommended).
pipeline_str = (
    f"filesrc location={INPUT_VIDEO} ! decodebin3 ! "
    "videoconvert ! "
    f"gvadetect model={MODEL}_openvino_model/{MODEL}.xml "
    "labels-file=coco_package_labels.txt "
    "device=GPU "
    "threshold=0.25 ! queue ! "
    "gvawatermark displ-cfg=show-roi=package ! "
    "videoconvert ! video/x-raw,format=I420 ! "
    "openh264enc bitrate=4000000 ! h264parse ! "
    "mp4mux ! filesink name=sink location=output_dlstreamer.mp4"
)
pipeline = Gst.parse_launch(pipeline_str)


def on_buffer(pad, info):
    buf = info.get_buffer()
    rmeta = GstAnalytics.buffer_get_analytics_relation_meta(buf)
    if rmeta is None:
        return Gst.PadProbeReturn.OK
    packages = []
    idx = 1
    while True:
        ok, od = rmeta.get_od_mtd(idx)
        if not ok:
            break
        label = GLib.quark_to_string(od.get_obj_type())
        if label in PACKAGE_LABELS:
            packages.append(label)
        idx += 1
    if packages:
        print(f"Packages detected: {len(packages)} ({', '.join(packages)})",
              flush=True)
    return Gst.PadProbeReturn.OK


sink = pipeline.get_by_name("sink")
sink.get_static_pad("sink").add_probe(Gst.PadProbeType.BUFFER, on_buffer)

pipeline.set_state(Gst.State.PLAYING)
bus = pipeline.get_bus()
bus.timed_pop_filtered(
    Gst.CLOCK_TIME_NONE,
    Gst.MessageType.EOS | Gst.MessageType.ERROR,
)
pipeline.set_state(Gst.State.NULL)

Device targets:

• device=GPU -- default in the sample code.
• device=CPU -- change device=GPU to device=CPU.
• device=NPU -- change device=GPU to device=NPU; use batch-size=1 and nireq=4 for best NPU utilization.

Expected Output

---

License

Licensed under the MIT License. See LICENSE for details.

References

• YOLO26 Documentation
• DETR Model Card (facebook/detr-resnet-50)
• End-to-End Object Detection with Transformers (DETR paper)
• OpenVINO Documentation
• Intel DLStreamer