Apppublic

johnodonovan/GPUtil

sourceHugging Faceupdated 7mo agoView on Hugging Face

1likes

App README

CloudPhysician GPU Capacity Planner

A Streamlit web application for estimating GPU requirements for AI-powered ICU patient monitoring systems.

Overview

CloudPhysician uses cameras in hospital ICUs to run various patient tracking AI models. This tool helps hardware managers understand how many GPUs will be needed to run their infrastructure based on:

—Number of hospital beds (camera feeds)
—Selected AI models (Mask2Former, YOLOv8, RF-DETR)
—GPU type selection
—Video resolution and frame rate requirements
—Inference frequency

Features

—Multi-model support: Estimate resources for running Mask2Former, YOLOv8, and RF-DETR simultaneously
—GPU comparison: Compare RTX 4080, RTX 5080, RTX 5090, and NVIDIA L40S
—Flexible inference modes: Continuous or interval-based processing
—Cost estimation: Total GPU cost and annual power consumption estimates
—Scaling projections: Visualize how requirements scale with bed count
—Password protected: Secure access for team members only

GPU Options

GPU	VRAM	FP16 TFLOPS	Price
RTX 4080	16 GB	48.7	$1,199
RTX 5080	16 GB	56.2	$999
RTX 5090	32 GB	104.8	$1,999
NVIDIA L40S	48 GB	91.6	$7,500

Technical Notes

—Inference times are estimated benchmarks based on typical model performance
—VRAM requirements are calculated per concurrent stream
—Power estimates use average electricity rates ($0.12/kWh)

License

Internal tool for CloudPhysician infrastructure planning.