[版本发布]OpenNCC百度定制版VCAM发布

近期，由百度美国研究院和EyeCloud共同研发的定制版OpenNCC开源AI摄像头VCAM正式上线GitHub, 感兴趣的朋友们可以在GitHub搜索hydra-vcam进行下载。

VCAM基于EyeCloud的OpenNCC开发，是百度Hydra AI项目的一款原型摄像头。

VCAM功能特点：

• 即插即用：将USB-C数据接口插入计算机即可开始开发创建自己的多功能AI摄像头。

• 标准模型：VCAM具有集成VPU芯片的优势，与百度Paddle和英特尔OpenVINO兼容。

• 多框架：VCAM支持Paddle、Caffe、ONNX、TensorFlow、MXNet等深度学习框架，便于用户友好地开发和使用。

• 高质量输出：VCAM已经过出厂视频调试，支持1920x1080或4K分辨率画质，支持YUV420、H.264、H265、MJPEG等视频格式输出。

• 二次开发：VCAM提供专用SDK开发工具包和相关技术文档，支持C/C++/Python语言。用户可以轻松调用相关API接口，启动摄像机参数设置、模型下载、输出视频参数设置，快速启动智能摄像机算法部署。VCAM支持Paddle和OpenVINO提供的官方模型，也支持用户自定义的算法模型。

VCAM SDK支持的操作系统：

• Ubuntu 16.04, Ubuntu 18.04
• Windows 10
• Raspberry Pi
• Arm Linux (需要工具链交叉编译)

VCAM SDK支持的编程语言：

• C/C++
• Python3.5, Python3.7

VCAM 摄像头型号：

• VCAM DK
• VCAM Lite
• VCAM USB

VCAM集成了的算法模型：

Model Category	Name	Brief Introduction
Object Classification	classification-fp16	ssd_mobilenet_v1_coco model can detect almost 90 objects
Face & Human Detection	face-detection-adas-0001-fp16	A face detector for driver monitoring and similar scenarios. The network features a default MobileNet backbone that includes depth-wise convolutions to reduce the amount of computation for the 3x3 convolution block
Face & Human Detection	face-detection-retail-0004-fp16	A face detector based on SqueezeNet light (half-channels) as a backbone with a single SSD for indoor/outdoor scenes shot by a front-facing camera
Face & Human Detection	face-person-detection-retail-0002-fp16	Pedestrian detector based on the backbone with hyper-feature + R-FCN for the Retail scenario
Face & Human Detection	person-detection-retail-0013-fp16	Pedestrian detector for the Retail scenario. It is based on MobileNetV2-like backbone that includes depth-wise convolutions to reduce the amount of computation for the 3x3 convolution block
Face & Human Detection	pedestrian-detection-adas-0002-fp16	Pedestrian detection network based on SSD framework with tuned MobileNet v1 as a feature extractor.
People, Vehicles & Bicycles Detection	person-vehicle-bike-detection-crossroad-0078-fp16	Person/Vehicle/Bike detector is based on SSD detection architecture, RMNet backbone, and learnable image downscale block (like person-vehicle-bike-detection-crossroad-0066, but with extra pooling)
People, Vehicles & Bicycles Detection	pedestrian-and-vehicle-detector-adas-0001-fp16	Pedestrian and vehicle detection network based on MobileNet v1.0 + SSD.
Vehicle Detection	vehicle-detection-adas-0002-fp16	A vehicle detection network based on an SSD framework with tuned MobileNet v1 as a feature extractor.
Mask	Detection	mask-detect-fp16 Mask detector.
License Plate Recognition	vehicle-license-plate-detection-barrier-0106-fp16	A MobileNetV2 + SSD-based vehicle and (Chinese) license plate detector for the “Barrier” use case.
Face Attributes	interactive_face_detection_demo	This demo executes four parallel infer requests for the Age/Gender Recognition, Head Pose Estimation, Emotions Recognition, and Facial Landmarks Detection networks that run simultaneously
Body Extraction	human-pose-estimation-0001-fp16	A multi-person 2D pose estimation network (based on the OpenPose approach) with tuned MobileNet v1 as a feature extractor.

VCAM运行机制

从模型训练环境到嵌入式部署，这是一项非常重要的任务，开发者需要掌握深度学习的框架，如常用的：Caffe*、TensorFlow*、MXNet*、Kaldi*等；此外，掌握部署的嵌入式平台非常重要，开发者需要了解平台性能、系统架构特点，然后结合平台特点优化培训模型框架，最后对嵌入式平台进行调优、移植和部署。

VCAM专注于深度学习模型的快速部署，与Intel OpenVINO工具兼容，适用于嵌入式图形和图像应用场景。它已在终端目标设备上完成了从2MP到20MP的不同分辨率传感器的集成，并在终端目标设备实现了专业级ISP的部署。OpenVINO优化的转换模型文件可以动态下载到终端VCAM摄像机，以实现深度学习模型的快速部署。VCAM设计了独立工作模式、混合开发模式和协同处理计算模式，以适应不同的工作应用场景。

欢迎对VCAM感兴趣的朋友前往GitHub项目页了解更多项目细节，也欢迎添加OpenNCC小助手微信好友（搜eyecloud666）加入OpenNCC开发技术交流微信群和工程师们一起交流。