Small Object Detection with YOLO: A Performance Analysis Across Model Versions and Hardware
Muhammad Fasih Tariq, Muhammad Azeem Javed · 2025
Read Small Object Detection with YOLO: A Performance Analysis Across Model Versions and Hardware by Muhammad Fasih Tariq, Muhammad Azeem Javed (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-09900.txt.
Description
Muhammad Fasih Tariq and Muhammad Azeem Javed test YOLO object detectors across versions and machines in a 2025 paper. YOLO models spot objects in images at high speed. Small objects are the hardest for them to catch. The team asked which version and setup trade speed and accuracy best.
They ran five YOLO versions, from v5 to v11. Tests ran on Intel and AMD chips and on graphics cards. Software stacks like ONNX and OpenVINO were compared. Objects took 1, 2.5, and 5 percent of each image in the checks. Results map the trade-offs for each setup. The guide helps users pick a model for their hardware limits.
Catalog record
| CREATOR | Muhammad Fasih Tariq, Muhammad Azeem Javed |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 12 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2504-09900