DeGirum

Building the future of edge AI takes more than cutting-edge technology. It takes an incredible team. The DeGirum crew recently stepped away from our screens to connect over great food and even better company. We fired up the grill, shared homemade dishes from around the world, and a few competitive frisbee games broke out. These moments remind us why we're passionate about fostering a collaborative community, not just among developers using our tools, but right here within our own team. Innovation thrives when people come together, whether it's solving edge AI challenges or perfecting BBQ techniques. Here's to the team behind DeGirum's mission to simplify edge AI development.

I just published a practical guide on building license plate detection and recognition systems using DeGirum PySDK with Hailo accelerators.   This walkthrough covers how to pipeline two AI models together: one for detecting license plates and another for reading the characters. The best part? It runs efficiently on edge devices without cloud dependency. What you'll learn: - Setting up model pipelining with DeGirum PySDK - Optimizing ANPR workflows for edge deployment - Working with Hailo hardware acceleration - Real-world implementation tips Whether you're working on parking management, access control, or traffic monitoring, this guide shows you how to get started quickly. Read the full guide: https://lnkd.in/g4GjX2yF

Building real-time license plate recognition is simpler than you think. DeGirum PySDK lets you chain detection and OCR models into one pipeline that works across multiple hardware platforms. Same code, different accelerators. Franklin M. breaks down the complete implementation in this technical walkthrough, covering everything from model loading to streaming inference. Applications include traffic monitoring, smart parking, toll automation, and fleet management. Read the guide below.

I just published a hands-on guide to building Multi-model, Multi-stream inference on Hailo using DeGirum PySDK + Tools. Three patterns I keep reusing on real projects: 1. 3 models × 3 streams (threaded) — each in its own thread for max concurrency 2. 3 models × 1 stream — fuse with a compound model (plus a manual fusion variant) 3. 3 models × 3 streams (single loop) — single-threaded loop with deterministic playback It’s all done in Python, using: - DeGirum AI Hub — ready-to-use models for HAILO8 / HAILO8L - ModelSpec — to define and load your models - predict_stream — to run inference on video streams - CombiningCompoundModel — to fuse multiple models together - Simple display helpers — to display the results 👉 Guide: https://lnkd.in/g2eZNgXH

Using DeGirum PySDK and running into “unable to connect to server” or “cloud model zoo doesn’t exists” errors? I put together a practical troubleshooting guide that starts at degirum.connect() and walks through the common failure points—env setup, model zoo config and access token issues. Check it out : https://lnkd.in/dm5JUDnm

We recently added support for Axelera AI METIS AI accelerator to our PySDK. We have planned a series of user guides to illustrate how edge AI applications can be developed with very simple code. Our first guide on getting started is now available on the DeGirum Community: https://lnkd.in/gG6MGepj

I’ve published a new step-by-step guide for engineers who want to get started quickly with DeGirum PySDK and Axelera AI's METIS hardware.   The guide covers how to:   - Install and verify DeGirum PySDK - Run inference in the cloud using DeGirum AI Hub - Run inference locally on METIS hardware - Visualize inference results for images and video   All examples are provided with concise Python code so you can begin running models with minimal setup.   If you're working with computer vision workloads and looking for a direct way to accelerate them on METIS hardware, this resource should be a good starting point. Check it out here: https://lnkd.in/gdSfgMT8

Deploying AI on the edge? Then you know the challenge: preserving accuracy while shrinking your model footprint. Some of our community members are sharing quantisation workflows that really deliver, and others, like the awesome team at DeGirum, are doing impressive things on their own platform. See their insights and share your own right here: 👉 https://lnkd.in/dqPtSUda #ModelOptimisation #Quantisation #EdgeAI

Excited to share my first DeGirum Community guide: Evaluating model accuracy after compilation with Hailo and PySDK   If you’re compiling models for Hailo accelerators, this walkthrough shows how to:   Run quick accuracy checks with degirum_tools Measure mAP for detection models and Top-K accuracy for classification Use simple Python snippets to validate results   Check it out here: https://lnkd.in/gF6dRKyp   Would love to hear your feedback or tips on how you approach accuracy evaluation!