ImageDenosing

This is a collection of image denosing 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)

模型	输入分辨率	AX650板端耗时
NAFNet	256×256	87ms
DnCNN	416×416	22ms
FFDNet	640×640	19.3ms
FastDVDnet	640×480	61.2ms

How to use

Download all files from this repository to the device

模型文件组织方式如下：
.
|-- pic
|   `-- noisy.png
|-- model_convert
|   |-- NAFNet.json
|   `-- axmodel
|       `-- NAFNet_1_3_256_256.axmodel
|-- python
|   |-- axmodel_infer.py
|   `-- onnx_infer.py
`-- res
    `-- axmodel_compare.png

Inference

Image Denoising

图片推理，执行命令python3 axmodel_infer.py:

(base) root@ax650:~/NAFNet# 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 115775d3
axmodel inference finished: axmodel_compare.png

推理结果样例：

Video Denoising

视频推理，执行命令python3 axmodel_video_infer.py:

(base) root@ax650:~/fastdvdnet# python3 axmodel_video_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 115775d3
providers: AxEngineExecutionProvider
inputs: [('noisy', [1, 15, 480, 640], 'float32'), ('noise_map', [1, 1, 480, 640], 'float32')]
outputs: [('denoised', [1, 3, 480, 640])]
video: 854x480 @ 25.0fps, 67 frames
axmodel input: 640x480
[OK] ./people-sunset.mp4 -> ./people-sunset_axmodel_side_by_side.mp4
     frames=67 time=17.57s fps=3.81

视频降噪样例：