一、deepstream配置组

1 应用组[application]

Key含义值类型示例平台
enable-perf-measurement表示命令行是否输出帧数布尔类型enable-perf-measurement=1dGPU,Jetson
per-measurement-interval-sec输出帧数的时间间隔(秒)无符合整形per-measurement-interval-sec=10dGPU,Jetosn
gie-kitti-output-dir以KITTI格式保存识别数据的路径字符串gie-kitti-output-dir=/primary-detectdGPU,Jetson
kitti-track-output-dir跟踪数据存储路径字符串kitti-track-output-dir=/trackerdGPU,Jetson

2 拼接显示[tiled-display]

Key含义值类型示例平台
enable表明是否拼接显示视频画面,当enable=2,[sink<%u>]组关键字link-to-demux=1,该[sink]组将会被连接到[source<%u>]整形enable=1dGPU,Jetson
rows拼接显示2D画面的行整形,>0row=5dGPU,Jetson
columns拼接显示2D画面的列整形,>0columns=6dGPU,Jetson
width以像素为单位,整个2D画面的宽整形,>0width=5dGPU,Jetson
height以像素为单位,整个2D画面的长整形,>0height=5dGPU,Jetson
nvbuf-memory-type元素输出buffers的类型, 0(nvbuf-mem-default),1(nvbuf-mem-cud-pinned),2(nvbuf-mem-cuda-device),3(nvbuf-mem-cuda-unified)整形[0,3]nvbuf-memory-type=3dGPU,Jetson

3 输入源[source]

Key含义值类型示例平台
enable使能或者不使能该源布尔类型enable=1dGPU,Jetson
typesource类型:1.Camera(V4L2) 2.URI 3.MultiURI 4.RTSP 5.Camera(CSI)(Jetson only)整形[1,5]type=1dGPU,Jetson
uriURI当type=2或者3,该选项有效,可以是file、HTTP URI、RTSP。字符串uri=file:///home/ubuntu/source.mp4 uri=http://127.0.0.1/source.mp4 uri=rtsp://127.0.0.1/source1 uri=file:///home/ubuntu/source_%d.mp4dGPU,Jetson
num-sources只有type=3的时候,该选项有效。表示一个源复用多少次整形>=0num-sources=2dGPU,Jetson
intra-decode-enable使能或不使能仅仅内部编码布尔类型intra-decode-enable=1dGPU,Jetson
num-extra-surfacesNumber of surfaces in addition to minimum decode surfaces given by the decoder. Can be used to manage the number of decoder output buffers in the pipeline.整形[0,24]num-exta-surfaces=5dGPU,Jetson
gpu-id该元素在那个GPU上运行整形>-0gpu-id=1dGPU
camera-id唯一的ID添加到元数据(Optional)整形>=0camera-id=2dGPU,Jetson
camera-width输入视频帧的宽度,仅type=1和5有效整形>=0camera-width=1920dGPU,Jetson
camera-height输入视频帧的高度,仅type=1和5有效整形>=0camera-height=1080dGPU,Jetson
camera-fps-n一秒,指定帧率的分子部分,仅type=1和5有效整形>=0camera-fps-n=30dGPU,Jetson
camera-fps-d一秒,指定帧率的分母部分,仅type=1和5有效整形>=0camera-fps-d=1dGPU,Jetson
camera-v4l2-dev-nodeV4L2设备节点,例如/dev/video<%u>整形>=0camera-v4l2-dev-node=1dGPU,Jetson
latency以毫秒为单位,平滑缓冲区大小,仅RTSP时候有效整形>=0latency=200dGPU,Jetson
camera-csi-sensor-id相机模块传感器ID。只有type=5时有效整形>=0camera-csi-sensor-id=1Jetson
drop-frame-interval丢帧的间隔。 例如,5表示解码器每隔五帧输出一次; 0表示没有丢帧整[0,30]drop-frame-interval=5dGPU,Jetson
nvbuf-memory-typebuffer内存类型 0 (cuda-pinned-mem): host/pinned memory allocated with cudaMallocHost(). 1 (cuda-device-mem): Device memory allocated with cudaMalloc(). 2 (cuda-unified-mem): Unified memory allocated with cudaMallocManaged().整形[0,2]nvbuf-memory-type=1dGPU
select-rtp-protocoltype=4有效,对于RTP传输协议选择,0:UDP + UDP Multicast + TCP, 1:仅用TCP整形0或4select-rtp-protocol=4dGPU, Jetson
rtsp-reconnect-interval-sectype-4有效,丢掉数据后重连间隔(秒)整形>=0rtsp-reconnect-interval-sec=60dGPU, Jetson
smart-record使能和不使能smart record布尔类型smart-record=1dGPU,Jetson
smart-rec-dir-path保存编码文件目录字符串smart-rec-dir-path=/home/nvidia/dGPU,Jetson
smart-rec-file-prefix记录文件前缀名字字符串smart-rec-file-prefix=Cam1dGPU,Jetson
smart-rec-video-cache以秒为单位,视频缓存大小整形>=0smart-rec-video-cache=20dGPU,Jetson
smart-rec-container保存视频格式,支持MP4和MKV整形[0,1]smart-rec-container=0dGPU,Jetson
smart-rec-start-timeNumber of seconds earlier from now to start the recording. E.g. if t0 is the current time and N is the start time in seconds that means recording will start from t0 – N. Obviously for it to work the video cache size must be greater than the N.整形>=0smart-rec-start-time=5dGPU,Jetson
smart-rec-default-durationIn case a Stop event is not generated. This parameter will ensure the recording is stopped after a predefined default duration.整形>=0smart-rec-default-duration=20dGPU,Jetson
smart-rec-durationDuration of recording in seconds.整形>=0smart-rec-duration=15dGPU,Jetson
smart-rec-intervalThis is the time interval in seconds for SR start / stop events generation.整形>=0smart-rec-interval=10dGPU,Jetson

smart部分有待测试

4 流合并组[streammux]

Key含义值类型示例平台
gpu-id多个GPU情况下,指定使用那个GPU整形>=0gpu-id=1dGPU
live-source告诉muxer这个source是实时流布尔类型live-source=0dGPU,Jetson
batch-sizeMuxer batch size整形>=0batch-size=4dGPU,Jetson
batched-push-timeout超时时间(ms)Integer, ≥−1batched-push-timeout=40000dGPU, Jetson
widthMuxer输出图像宽度Integer, >0width=1280dGPU, Jetson
heightMuxer输出图像高度Integer, >0height=720dGPU, Jetson
enable-paddingIndicates whether to maintain source aspect ratio when scaling by adding black bands.布尔类型enable-padding=0dGPU,Jetson
nvbuf-memory-typeType of CUDA memory the element is to allocate for output buffers. 0 (nvbuf-mem-default, a platform-specific default. 1 (nvbuf-mem-cuda-pinned): pinned/host CUDA memory. 2 (nvbuf-mem-cuda-device): Device CUDA memory. 3 (nvbuf-mem-cuda-unified): Unified CUDA memory.For dGPU: All values are valid.For Jetson: Only 0 (zero) is valid.整形[0,3]nvbuf-memory-type=3dGPU
attach-sys-ts-as-ntp1表示添加系统时间戳到Frame元数据。2表示添加来自rstp源时间戳到元数据布尔类型attach-sys-ts-as-ntp=0dGPU, Jetson
config-file-path仅对新创建的streammux有效,更多信息参考“New Gst-nvstreammux”字符串config-file-path=config_mux_source30.txtdGPU, Jetson

5 一级GIE [primary-gie]和二级GIE [secondary-gie1]

Key含义值类型示例平台/GIE
enable表示是否使用该组布尔类型enable=1dGPU, Jetson、Both GIEs
gie-unique-id该部分唯一ID整形>0gie-unique-id=2Both
gpu-idGPU to be used by the element in case of multiple GPUs.整形≥0gpu-id=1dGPU, Jetson、Both GIEs
mode-engine-file模型路径字符串model-engine-file=…/resnet10. caffemodel_b4_int8.engineBoth GIEs
nvbuf-memory-typeType of CUDA memory element is to allocate for output buffers. 0 (nvbuf-mem-default): a platform-specific default. 1 (nvbuf-mem-cuda-pinned): pinned/host CUDA memory. 2 (nvbuf-mem-cuda-device): Device CUDA memory. 3 (nvbuf-mem-cuda-unified): Unified CUDA memory. For dGPU: All values are valid.For Jetson: Only 0 (zero) is valid.整形[0,3]nvbuf-memory-type=3dGPU, Jetson、Primary GIE
config-file指定推理插件属性,一定被定义在[property]组,更多变量信息“Gst-nvinfer File Configuration Specifications”在Deepstream 插件手册字符串config-file=…/resnet.txtdGPU, Jetson、Both GIE
batch-size一个batch有几个源去推理整形>0batch-size=2dGPU, Jetson、Both GIEs
interval间隔多少帧推理整形>0batch-size=2dGPU, Jetson、Primary GIE
bbox-border-color识别对象框颜色浮点型RGBA[0,1]bbox-border-color2=1;0;0;1(Red for class-id 2)dGPU, Jetson、Both GIEs
bbox-bg-color识别对象框背景颜色浮点型RGBA[0,1]bbox-bg-color3=0;1;0;0.3=1;0;0;1(Green for class-id 3)dGPU, Jetson、Both GIEs
operate-on-gie-id一个唯一的GIE ID,这个ID会被添加到Frame元数据整形>0operate-on-gie-id=1dGPU, Jetson Secondary GIE
operate-on-class-idsClass IDs of the parent GIE on which this GIE must operate. The parent GIE is specified using operate-on-gie-id.Semicolon separated integer arrayoperate-on-class-ids=1;2(operate on objects with class IDs 1, 2 generated by parent GIE)
infer-raw-output-dir指定文件夹保存未处理buffer.字符串infer-raw-output-dir=/home/ubuntu/infer_raw_outdGPU, Jetson、Both GIEs
labelfile-path标签(label)字符串labelfile-path=…/labels.txtdGPU, Jetson、Both GIEs
plugin-type推理插件类型 0:nvinfer(TensorRt) 1:nvinferserver(Triton inference server)整形[0,1]plugin-type=1dGPU, Jetson、Both GIEs

6 跟踪组[tracker]

Key含义值类型示例平台
enable使能或不使能该组布尔类型enable=1dGPU, Jetson
tracker-width跟踪器宽度Integer, ≥0tracker-width=960dGPU, Jetson
tracker-height跟踪器高度整形Integer, ≥0tracker-height=540dGPU, Jetson
gpu-id多个GPU情况下,指定GPU整形Integer, ≥0gpu-id=1dGPU, Jetson
ll-config-file跟踪器配置文件字符串ll-config-file=iou_config.txtdGPU, Jetson
ll-lib-file跟踪器库文件字符串ll-lib-file=/usr/local/deepstream/libnvds_mot_iou.sodGPU, Jetson
enable-batch-process使能多个流批处理布尔类型enable-batch-process=1dGPU, Jetson
enable-past-frameEnables reporting past-frame data布尔类型enable-past-frame=1dGPU, Jetson

  英文未翻译部分,该变量还没有测试,不能确定具体效果,日后各插件用到时,再做补充。

参考

参考1:Deepstream Configuration Groups
参考2:KITTI数据集–参数

  • 0
    点赞
  • 8
    收藏
    觉得还不错? 一键收藏
  • 0
    评论
评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值