2025.12.20更新：
压缩numpy，opencv-python，visiong库大小，降低内存需求，内存有限的G2版本依旧建议使用低分辨率；

使用：
1.将附件中的numpy_cv2_visiong.zip中的内容解压到/usr/lib/python3.11/site-packages/即可（注意剩余空间是否满足所需）
2.联合使用示例可见压缩包内的test.py文件，旧的演示视频：https://www.bilibili.com/video/BV1HdeEz4EYU

说明：
1.移植的numpy的版本是1.24.3，为了实现对 uclibc 环境的兼容，禁用了部分依赖 ...

My project implements the ability to call the ISP to modify saturation, sharpness, etc. Although it's mainly a Python library, I also provide a C++ interface: https://github.com/yiex/visiong
You can either directly call my library, or look at the specific implementation in my code (visiong/src ...

简单修改寄存器和iqfile，未走完整标定流程，画面基本正常。
sc3336.ko替换/oem/usr/ko/sc3336.ko
sc3336_CMK-OT2119-PC1_30IRC-F16.json替换/oem/usr/share/iqfiles/sc3336_CMK-OT2119-PC1_30IRC-F16.json

我已将visiong库开源, https://github.com/yiex/visiong/

Hi Salieri,
I confirmed that visiong works perfectly (YOLOv11n, camera, NPU, Vigide handshake OK).
The only blocking issue is that DisplayUDP destination IP is hardcoded to 172.32.0.100 in visiong.so.
Even when setting:
DISPLAY_UDP_IP=xxx.xxx.xxx.xxx
DISPLAY_UDP_PORT=xxxx
the runtime still prints ...

https://github.com/yiex/visiong/
可以降低使用门槛

有点想不开了，这种IPC芯片的主要针对CNN设计，对Transformer里常用的MatMul、LayerNorm这些关键算子都不支持

Hi Salieri,
I confirmed that visiong works perfectly (YOLOv11n, camera, NPU, Vigide handshake OK).
The only blocking issue is that DisplayUDP destination IP is hardcoded to 172.32.0.100 in visiong.so.
Even when setting:
DISPLAY_UDP_IP=xxx.xxx.xxx.xxx
DISPLAY_UDP_PORT=xxxx
the runtime still prints ...

1 MB = 8 Mb,'Byte' 'bit'

jjw wrote: ↑2025-11-27 7:57 我测试了彩条图，LVGL和OpencvMobile显示基本一致，但与原始图片显示颜色错误。
color3.jpg
下面是Lvgl显示
color1.jpg
下面是用摄像头对着彩条色拍照，然后在OPencv显示的
color2.jpg

屏幕颜色反向，加BGR排列；你需要反转屏幕颜色，并对调B和R色相