LowLightImageEnhancement

This is a collection of Low-light Enhancement algorithms, models have been converted to run on the Axera NPU using w8a8 quantization.

This model has been optimized with the following LoRA:

Compatible with Pulsar2 version: 6.0 115775d3

Convert tools links:

For those who are interested in model conversion, you can try to export axmodel through

• Pulsar2 Link, How to Convert ONNX to axmodel

Support Platform

• AX650
  • M4N-Dock(爱芯派Pro)
  • M.2 Accelerator card

性能基准测试 (Performance Benchmark)

Model	Input Size	Inference Time
Zero-DCE	256×256	2.6ms
Zero-DCE++	512×512	12.7ms
SCI	600×400	1.2ms
RetinexFormer	224×224	35ms

How to use

Download all files from this repository to the device

模型文件组织方式如下：
.
|-- model_convert
|   |-- SCI.json
|   `-- axmodel
|       `-- SCI_TPAMI_600_400.axmodel
|-- pic
|   |-- 00001.png
|   |-- 00051.png
|   |-- 00079.png
|   |-- 00091.png
|   |-- 2062.jpg
|   |-- 2064.jpg
|   |-- 3008.jpg
|   |-- 3018.jpg
|   |-- 3020.jpg
|   `-- NPE_71.png
|-- python
|   |-- axmodel_infer.py
|   `-- onnx_infer.py
`-- res
    `-- axmodel_res.png

Inference

图片推理，执行命令python3 axmodel_infer.py:

(base) root@ax650:~/SCI# python3 axmodel_infer.py
[INFO] Available providers:  ['AxEngineExecutionProvider', 'AXCLRTExecutionProvider']
[INFO] Using provider: AxEngineExecutionProvider
[INFO] Chip type: ChipType.MC50
[INFO] VNPU type: VNPUType.DISABLED
[INFO] Engine version: 2.12.0s
[INFO] Model type: 2 (triple core)
[INFO] Compiler version: 6.0-dirty 115775d3-dirty
Saved comparison → ./axmodel_res.png
Input: (600, 400)  →  model input: (600,400)  →  restored: (600, 400)

推理结果样例：