# I Compared 3 Lightweight CV Models for Edge Deployment — retained record

> Retained reddit record from reddit:computervision.

Canonical: https://highsignal.app/data/records/1ad07fa815ca03d7

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:computervision)  published 2026-09-19 19:37 ingested 2026-09-20 02:36

[original source ↗](https://www.reddit.com/r/computervision/comments/1wkvfdq/i_compared_3_lightweight_cv_models_for_edge/)

## retained text



```
I’ve been playing around with popular computer vision models to find out which ones are best optimized for edge devices, so I decided to compare **YOLO 26 Nano, YOLO 12 Nano, and RF-DETR Small** under the same setup.

Instead of looking only at accuracy, I wanted to see what actually happens during inference:

* FPS / inference speed
* P50 and P95 latency
* VRAM usage
* RAM usage
* Detection count
* Performance across video frames

For this comparison, I’m only testing the **Nano variants or their closest equivalent small models**, since the goal is to evaluate models suitable for edge-device deployment.

I’ve shared the notebook/code and the full benchmarking process here:

Notebook: [LINK](https://github.com/Labellerr/Hands-On-Learning-in-Computer-Vision/blob/main/Experiments/vision_models_nano_variant_comparison.ipynb)

Video: [LINK ](https://youtu.be/qPKYO6WvNiM)
```



document fetched 2026-09-20 00:22 · canonical https://reddit.com/r/computervision/comments/1wkvfdq/i_compared_3_lightweight_cv_models_for_edge
