Downloads · 30 days
0
resnet50
AXERA-TECH/resnet50
resnet50 is a image classification model from AXERA-TECH. Use it when you need a label for an image. It is set up for axengine. The card lists the license as apache-2.0.
ResNet50 图像分类模型，已编译为 AX650 芯片的 AXMODEL 格式。
Access
Public
Updated Jul 15, 2026
Repo size
26.9 MB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.axmodel26.9 MB · 100%
At a glance
- Task
- Image Classification
- Library
- axengine
- License
- apache-2.0
- Access
- Public
- Created
- Jul 15, 2026
- Updated
- Jul 15, 2026
- SHA
- e3620065
Model card
From the Hugging Face model README
ResNet50 ImageNet Classification (AX650)
ResNet50 图像分类模型，已编译为 AX650 芯片的 AXMODEL 格式。
模型概述
- 任务: 图像分类 (ImageNet 1000类)
- 芯片: AX650 / AX650N
- 输入: [1, 3, 224, 224], float32, NCHW
- 预处理: (input/255 - mean) / std, mean=[0.485,0.456,0.406], std=[0.229,0.224,0.225]
- 输出: [1, 1000], float32, ImageNet 类别 logits
目录说明
| 目录 | 用途 |
|---|---|
models/ | AXMODEL 模型文件 + 元信息 |
python/ | Python SDK (基于 pyaxengine) |
model_convert/ | 从零复现模型转换 |
快速开始
路径 A: 直接用 AXMODEL 推理
环境安装:
# 安装 pyaxengine
pip install pyaxengine
# 或从源码: git clone https://github.com/AXERA-TECH/pyaxengine && cd pyaxengine && pip install .
# 安装依赖
cd python
pip install -r requirements.txt
运行推理:
cd python
# 确保 LD_LIBRARY_PATH 包含 AX 运行时库
export LD_LIBRARY_PATH=/soc/lib:$LD_LIBRARY_PATH
python example.py <image.jpg>
路径 B: 从零复现
性能摘要
| 指标 | 值 |
|---|---|
| MACs | 4.09G |
| AXMODEL 大小 | 26 MB |
| ONNX 大小 | 102 MB |
| 压缩比 | 3.94x |
| 仿真 Cosine | 0.9997 ± 0.0001 |
| 板端 Python 延迟 | ~1600 ms |
已知限制
- 输入分辨率固定为 224x224
- Batch size 固定为 1
- 仅 AX650/AX650N 芯片
- Task
- Image Classification
- Library
- axengine
- License
- apache-2.0
- Languages
- en
- Created
- Jul 15, 2026
- Updated
- Jul 15, 2026