System Requirements#
This section summarizes the recommended hardware, software, and network requirements for running the multi‑modal patient monitoring application.
Host Operating System#
Ubuntu 22.04 LTS (recommended and validated).
Other recent 64‑bit Linux distributions may work, but are not fully validated.
Hardware Requirements#
CPU:
4 physical cores (8 threads) or more recommended.
x86_64 architecture with support for AVX2.
System Memory (RAM):
Minimum: 16 GB.
Recommended: 32 GB for smoother multi‑service operation and development work.
Storage:
Minimum free disk space: 30 GB.
Recommended: 50 GB+ to accommodate Docker images, models, logs, and metrics.
Graphics / Accelerators:
Required: Intel CPU.
Optional (recommended for full experience):
Intel integrated GPU supported by Intel® Graphics Compute Runtime.
Intel NPU supported by the linux‑npu‑driver stack.
The host must expose GPU and NPU devices to Docker, for example:
/dev/dri(GPU)/dev/accel/accel0(NPU)
Software Requirements#
Docker and Container Runtime:
Docker Engine 24.x or newer.
Docker Compose v2 (integrated as
docker compose) or compatible compose plugin.Ability to run containers with:
--privileged(for metrics‑collector).Device mappings for GPU/NPU (for rPPG and metrics‑collector).
Python (for helper scripts and tools):
Python 3.10 or newer recommended.
Used primarily for asset preparation scripts and local tooling; application containers include their own Python runtimes (for example, Python 3.12 in the rPPG service image).
Git and Make:
gitfor cloning the repository and managing submodules.maketo run provided automation targets (e.g.,make run,make init-mdpnp).
AI Models and Workloads#
The application bundles several AI workloads, each with its own model(s) and inputs/outputs:
RPPG (Remote Photoplethysmography) Workload:
Model: MTTS‑CAN (Multi‑Task Temporal Shift Convolutional Attention Network) converted to OpenVINO IR (
/models/rppg/mtts_can.xml).Input: Facial video frames (RGB) from the shared
videosvolume.Output: Pulse and respiration waveforms, heart rate (HR) in BPM, and respiratory rate (RR) in BrPM.
Target devices: Intel CPU, Intel integrated GPU, or Intel NPU via OpenVINO (
RPPG_DEVICE).
3D‑Pose Estimation Workload:
Model:
human-pose-estimation-3d-0001from Open Model Zoo, converted to OpenVINO IR (/models/3d-pose/human-pose-estimation-3d-0001.xml).Input: RGB video of a person in motion (
face-demographics-walking.mp4under/videos/3d-posehas been provided for demonstration purposes).Output: 3D human keypoints and pose estimation, streamed to the aggregator for visualization.
Target devices: Intel CPU and GPU via OpenVINO.
AI‑ECG Workload:
Models: OpenVINO IR models for ECG rhythm classification located under
/models/ai-ecg, for example:ecg_8960_ir10_fp16.xmlecg_17920_ir10_fp16.xml
Input: Preprocessed multi‑lead ECG time‑series segments of supported lengths (e.g., 8960 or 17920 samples).
Output: Rhythm classification labels (e.g., Normal sinus rhythm, Atrial Fibrillation, Other rhythm, or Too noisy to classify) with associated waveforms and timings.
Target devices: Intel CPU, GPU, or other OpenVINO‑supported devices configured via
ECG_DEVICE.
Network and Proxy#
Network Access:
Local network connectivity to access the UI (default:
http://<HOST_IP>:3000).Optional outbound internet access to download Docker base images, models, and assets (if not pre‑cached).
Proxy Support (optional):
If your environment uses HTTP/HTTPS proxies, configure:
HTTP_PROXY,HTTPS_PROXY,NO_PROXYin the shell before runningmake.
Permissions#
Ability to run Docker as a user in the
dockergroup or withsudo.Sufficient permissions to access device nodes for GPU and NPU (typically via membership in groups such as
videoor via explicitdevicesconfiguration in Docker Compose).
Browser Requirements#
Modern web browser (Chrome, Edge, or Firefox) to access the UI dashboard.
JavaScript enabled.
These requirements are intended for development and evaluation environments. For any production‑like deployment, you should also consider additional factors such as security hardening, monitoring, backup, and resource isolation.