---
library_name: pytorch
---

![yolox_logo](resource/YOLOX.png)

YOLOX modernizes one-stage object detection by adopting an anchor-free design and decoupled classification and regression heads, improving both accuracy and convergence speed.

Original paper: [YOLOX: Exceeding YOLO Series in 2021](https://arxiv.org/abs/2107.08430)

# YOLOX-S

YOLOX-S (Small) is a lightweight variant optimized for fast inference while maintaining competitive detection accuracy. It is well suited for real-time object detection in applications such as video analytics, robotics, and edge deployment where low latency is critical.

Model Configuration:
- Reference implementation: [YOLOX](https://github.com/Megvii-BaseDetection/YOLOX)
- Original Weight: [YOLOX_S_Weights.COCO2017](https://github.com/Megvii-BaseDetection/YOLOX/releases/download/0.1.1rc0/yolox_s.pth)
- Resolution: 3x640x640
- Support Cooper version:
    - Cooper SDK: [2.5.2]
    - Cooper Foundry: [2.2]

| Model | Device | Model Link |
| :-----: | :-----: | :-----: |
| YOLOX-s | N1-655 | [Model_Link](https://huggingface.co/Ambarella/YOLOX/blob/main/n1-655_yolox_s.bin) |
| YOLOX-s | CV72 | [Model_Link](https://huggingface.co/Ambarella/YOLOX/blob/main/cv72_yolox_s.bin) |
| YOLOX-s | CV75 | [Model_Link](https://huggingface.co/Ambarella/YOLOX/blob/main/cv75_yolox_s.bin) |