备份最新边缘设备程序
This commit is contained in:
BBIT-Kai
+25
@@ -0,0 +1,25 @@
|
||||
[property]
|
||||
enable=1
|
||||
#Width height used for configuration to which below configs are configured
|
||||
config-width=1920
|
||||
config-height=1080
|
||||
osd-mode=2
|
||||
display-font-size=12
|
||||
|
||||
[line-crossing-stream-0]
|
||||
enable=1
|
||||
# 箭头尾部、箭头头部、横线开始、横线结束、550;550;700;800;200;650;1700;650 # 测试视频
|
||||
line-crossing-LineA=950;400;1050;600;800;460;1060;460
|
||||
line-crossing-LineB=1040;630;1200;900;870;750;1280;720
|
||||
class-id=0
|
||||
extended=0
|
||||
# mode=loose、strict、balanced
|
||||
mode=loose
|
||||
|
||||
[line-crossing-stream-1]
|
||||
enable=1
|
||||
line-crossing-LineA=930;210;850;430;780;330;1000;330
|
||||
line-crossing-LineB=750;550;620;900;340;680;990;700
|
||||
class-id=0
|
||||
extended=0
|
||||
mode=loose
|
||||
+61
@@ -0,0 +1,61 @@
|
||||
BaseConfig:
|
||||
minDetectorConfidence: 0.0430 # 如果检测器检测框的置信度低于此值，则不会被用于跟踪
|
||||
|
||||
TargetManagement:
|
||||
enableBboxUnClipping: 1 # 如果检测框可能被图像边界裁剪，则解除裁剪
|
||||
preserveStreamUpdateOrder: 0 # 分配新目标ID时，是否保持输入流顺序，以在多次运行中保持目标ID确定性
|
||||
maxTargetsPerStream: 150 # 每个流可跟踪的最大目标数。建议设置大于10。注意：此值应包括以阴影模式跟踪的目标。最大值取决于GPU内存容量
|
||||
|
||||
# [目标创建与终止策略]
|
||||
minIouDiff4NewTarget: 0.7418 # 如果新检测到的对象与已有目标的IOU高于此阈值，则丢弃该新对象
|
||||
minTrackerConfidence: 0.4009 # 如果对象跟踪器的置信度低于此值，则以阴影模式跟踪。有效范围：[0.0, 1.0]
|
||||
probationAge: 2 # 目标年龄超过此值，则视为有效目标
|
||||
maxShadowTrackingAge: 51 # 阴影跟踪的最大长度。如果阴影跟踪年龄超过此值，跟踪器将被终止
|
||||
earlyTerminationAge: 1 # 如果阴影跟踪年龄在试验期达到此阈值，则目标会提前终止
|
||||
|
||||
TrajectoryManagement:
|
||||
useUniqueID: 0 # 分配跟踪器ID时是否使用64位长唯一ID
|
||||
|
||||
DataAssociator:
|
||||
dataAssociatorType: 0 # 数据关联器类型 { DEFAULT=0 }
|
||||
associationMatcherType: 1 # 匹配算法类型 { GREEDY=0, CASCADED=1 }
|
||||
checkClassMatch: 1 # 是否只将同类对象关联，默认值：true
|
||||
|
||||
# [关联度指标：有效候选阈值]
|
||||
minMatchingScore4Overall: 0.4290 # 总匹配分数最小值
|
||||
minMatchingScore4SizeSimilarity: 0.3627 # 检测框尺寸相似度最小分数
|
||||
minMatchingScore4Iou: 0.2575 # IOU最小分数
|
||||
minMatchingScore4VisualSimilarity: 0.5356 # 视觉相似度最小分数
|
||||
|
||||
# [关联度指标：权重]
|
||||
matchingScoreWeight4VisualSimilarity: 0.3370 # 视觉相似度权重（相关响应比率）
|
||||
matchingScoreWeight4SizeSimilarity: 0.4354 # 尺寸相似度权重
|
||||
matchingScoreWeight4Iou: 0.3656 # IOU权重
|
||||
|
||||
# [关联度指标：试探性检测] 仅使用IOU相似度进行试探性检测匹配
|
||||
tentativeDetectorConfidence: 0.2008 # 如果检测置信度低于此值但高于minDetectorConfidence，则视为试探性检测
|
||||
minMatchingScore4TentativeIou: 0.5296 # 匹配目标与试探性检测的最小IOU阈值
|
||||
|
||||
StateEstimator:
|
||||
stateEstimatorType: 1 # 状态估计器类型 { DUMMY=0, SIMPLE=1, REGULAR=2 }
|
||||
|
||||
# [动态建模]
|
||||
processNoiseVar4Loc: 1.5110 # 检测框中心的过程噪声方差
|
||||
processNoiseVar4Size: 1.3159 # 检测框尺寸的过程噪声方差
|
||||
processNoiseVar4Vel: 0.0300 # 速度的过程噪声方差
|
||||
measurementNoiseVar4Detector: 3.0283 # 检测器检测的测量噪声方差
|
||||
measurementNoiseVar4Tracker: 8.1505 # 跟踪器定位的测量噪声方差
|
||||
|
||||
VisualTracker:
|
||||
visualTrackerType: 1 # 视觉跟踪器类型 { DUMMY=0, NvDCF=1 }
|
||||
|
||||
# [NvDCF：特征提取]
|
||||
useColorNames: 1 # 是否使用ColorNames特征
|
||||
useHog: 0 # 是否使用HOG（方向梯度直方图）特征
|
||||
featureImgSizeLevel: 2 # 特征图尺寸等级。有效范围：{1, 2, 3, 4, 5}，从最小到最大
|
||||
featureFocusOffsetFactor_y: -0.2000 # 汉宁窗口中心相对于特征高度的偏移量。中心在垂直方向移动 (featureFocusOffsetFactor_y * featureMatSize.height)
|
||||
|
||||
# [NvDCF：相关滤波器]
|
||||
filterLr: 0.0750 # DCF滤波器指数移动平均学习率，有效范围：[0.0, 1.0]
|
||||
filterChannelWeightsLr: 0.1000 # 特征通道权重的学习率，有效范围：[0.0, 1.0]
|
||||
gaussianSigma: 0.7500 # 创建DCF滤波器时期望响应的高斯标准差（像素）
|
||||
Vendored
+55
@@ -0,0 +1,55 @@
|
||||
source-list:
|
||||
- main: rtsp://admin:cda1b2c3@192.168.1.6:554/Streaming/Channels/101 # 正1
|
||||
side: rtsp://admin:cda1b2c3@192.168.1.4:554/Streaming/Channels/101 # 辅2
|
||||
- main: rtsp://admin:cda1b2c3@192.168.1.7:554/Streaming/Channels/101 # 正2
|
||||
side: rtsp://admin:cda1b2c3@192.168.1.8:554/Streaming/Channels/101 # 辅1
|
||||
|
||||
# 测试环境
|
||||
# - main: file:///opt/nvidia/deepstream/deepstream/sources/apps/sample_apps/sentinel/dist/source/traffic_test.mp4
|
||||
# side: rtsp://admin:cda1b2c3@10.0.4.41:554/Streaming/Channels/101
|
||||
|
||||
vehicle_type:
|
||||
# 0：不限制 1：轿跑车 2：大型车辆 3：轿车 4：SUV 5：卡车 6：面包车/厢型车
|
||||
# 7：大型车辆 + 卡车（生产）
|
||||
type: 0
|
||||
|
||||
output:
|
||||
## 1:file ouput 2:fake output 3:eglsink output 4:rtsp output
|
||||
type: 4
|
||||
## 0: H264 encoder 1:H265 encoder
|
||||
enc: 0
|
||||
## encoder type 0=Hardware 1=Software
|
||||
enc-type: 0
|
||||
bitrate: 4000000
|
||||
##The file name without suffix
|
||||
filename: test
|
||||
|
||||
primary-gie:
|
||||
#0:nvinfer, 1:nvinfeserver
|
||||
plugin-type: 0
|
||||
##For car detection
|
||||
config-file-path: pgie_0_traffic_cam_net.yml
|
||||
unique-id: 1
|
||||
|
||||
#If there is ROI
|
||||
analytics:
|
||||
enable: 1
|
||||
config-file: config_nvdsanalytics.txt
|
||||
|
||||
secondary-gie1:
|
||||
unique-id: 2
|
||||
#0:nvinfer, 1:nvinfeserver
|
||||
plugin-type: 0
|
||||
config-file-path: sgie_1_vehicle_type.yml
|
||||
process-mode: 2
|
||||
|
||||
tracker:
|
||||
enable: 1
|
||||
tracker-width: 640
|
||||
tracker-height: 384
|
||||
gpu-id: 0
|
||||
ll-lib-file: /opt/nvidia/deepstream/deepstream/lib/libnvds_nvmultiobjecttracker.so
|
||||
ll-config-file: config_tracker.yml
|
||||
enable-batch-process: 1
|
||||
|
||||
|
||||
Vendored
+55
@@ -0,0 +1,55 @@
|
||||
source-list:
|
||||
- main: rtsp://admin:cda1b2c3@192.168.1.6:554/Streaming/Channels/101 # 正1
|
||||
side: rtsp://admin:cda1b2c3@192.168.1.4:554/Streaming/Channels/101 # 辅2
|
||||
- main: rtsp://admin:cda1b2c3@192.168.1.7:554/Streaming/Channels/101 # 正2
|
||||
side: rtsp://admin:cda1b2c3@192.168.1.8:554/Streaming/Channels/101 # 辅1
|
||||
|
||||
# 测试环境
|
||||
# - main: file:///opt/nvidia/deepstream/deepstream/sources/apps/sample_apps/sentinel/dist/source/traffic_test.mp4
|
||||
# side: rtsp://admin:cda1b2c3@10.0.4.41:554/Streaming/Channels/101
|
||||
|
||||
vehicle_type:
|
||||
# 0：不限制 1：轿跑车 2：大型车辆 3：轿车 4：SUV 5：卡车 6：面包车/厢型车
|
||||
# 7：大型车辆 + 卡车（生产）
|
||||
type: 7
|
||||
|
||||
output:
|
||||
## 1:file ouput 2:fake output 3:eglsink output 4:rtsp output
|
||||
type: 2
|
||||
## 0: H264 encoder 1:H265 encoder
|
||||
enc: 0
|
||||
## encoder type 0=Hardware 1=Software
|
||||
enc-type: 0
|
||||
bitrate: 6000000
|
||||
##The file name without suffix
|
||||
filename: test
|
||||
|
||||
primary-gie:
|
||||
#0:nvinfer, 1:nvinfeserver
|
||||
plugin-type: 0
|
||||
##For car detection
|
||||
config-file-path: pgie_0_traffic_cam_net.yml
|
||||
unique-id: 1
|
||||
|
||||
#If there is ROI
|
||||
analytics:
|
||||
enable: 1
|
||||
config-file: config_nvdsanalytics.txt
|
||||
|
||||
secondary-gie1:
|
||||
unique-id: 2
|
||||
#0:nvinfer, 1:nvinfeserver
|
||||
plugin-type: 0
|
||||
config-file-path: sgie_1_vehicle_type.yml
|
||||
process-mode: 2
|
||||
|
||||
tracker:
|
||||
enable: 1
|
||||
tracker-width: 640
|
||||
tracker-height: 384
|
||||
gpu-id: 0
|
||||
ll-lib-file: /opt/nvidia/deepstream/deepstream/lib/libnvds_nvmultiobjecttracker.so
|
||||
ll-config-file: config_tracker.yml
|
||||
enable-batch-process: 1
|
||||
|
||||
|
||||
+49
@@ -0,0 +1,49 @@
|
||||
property:
|
||||
# 官方文档中的配置=============================
|
||||
gpu-id: 0
|
||||
infer-dims: 3;544;960
|
||||
net-scale-factor: 0.00392156862745098
|
||||
tlt-model-key: tlt_encode
|
||||
# 模型网络类型 0: Detector 1: Classifier 2: Segmentation 3: Instance Segmentation
|
||||
network-type: 0
|
||||
num-detected-classes: 4
|
||||
# 模型颜色格式 0：RGB 1：BGR 2：灰度
|
||||
model-color-format: 0
|
||||
# 保持宽高比 0：不保持 1：保持
|
||||
maintain-aspect-ratio: 0
|
||||
# 是否输出张量元数据 0：否 1：是
|
||||
output-tensor-meta: 0
|
||||
|
||||
# 文件路径配置=============================
|
||||
onnx-file: ../models/traffic_cam_net/resnet18_trafficcamnet_pruned.onnx
|
||||
labelfile-path: ../models/traffic_cam_net/labels.txt
|
||||
model-engine-file: ../models/traffic_cam_net/resnet18_trafficcamnet_pruned.onnx_b4_gpu0_fp16.engine
|
||||
|
||||
gie-unique-id: 1
|
||||
batch-size: 4
|
||||
# 推理模式配置 1：整帧 2：上游ROI （1：主推理 2：次推理（基于主推理））
|
||||
process-mode: 1
|
||||
## 0=FP32, 1=INT8, 2=FP16 mode
|
||||
network-mode: 2
|
||||
# 推理间隔 0：每帧 1：每1帧 2：每2帧
|
||||
interval: 0
|
||||
output-blob-names: output_cov/Sigmoid:0;output_bbox/BiasAdd:0
|
||||
# 集群模式 0: OpenCV groupRectangles() 1: DBSCAN 2: Non Maximum Suppression 3: DBSCAN + NMS Hybrid 4: No clustering
|
||||
cluster-mode: 2
|
||||
# 计算硬件 0: Platform default – GPU (dGPU), VIC (Jetson) 1: GPU 2: VIC (Jetson only)
|
||||
scaling-compute-hw: 0
|
||||
|
||||
# 设置具体类别NMS等参数
|
||||
class-attrs-all:
|
||||
# 保留的最多对象数
|
||||
topk: 20
|
||||
# 两个方案之间的最大 IOU 分数，超过该分数后，置信度较低的方案将被拒绝。
|
||||
nms-iou-threshold: 0.5
|
||||
# 预聚类阈值，范围为0到1。较高的值会导致更多的边界框被聚类在一起。
|
||||
pre-cluster-threshold: 0.2
|
||||
|
||||
## Per class configurations
|
||||
class-attrs-0:
|
||||
topk: 20
|
||||
nms-iou-threshold: 0.6
|
||||
pre-cluster-threshold: 0.4
|
||||
+28
@@ -0,0 +1,28 @@
|
||||
property:
|
||||
gpu-id: 0
|
||||
gie-unique-id: 2
|
||||
net-scale-factor: 1
|
||||
scaling-compute-hw: 0
|
||||
network-type: 1
|
||||
infer-dims: 3;224;224
|
||||
onnx-file: ../models/vehicle_type_net/resnet18_pruned.onnx
|
||||
model-engine-file: ../models/vehicle_type_net/resnet18_pruned.onnx_b4_gpu0_fp16.engine
|
||||
# model-engine-file: ../models/vehicle_type_net/resnet18_pruned.onnx_b4_gpu0_int8.engine
|
||||
labelfile-path: ../models/vehicle_type_net/labels.txt
|
||||
int8-calib-file: ../models/vehicle_type_net/resnet18_pruned_int8.txt
|
||||
batch-size: 4
|
||||
network-mode: 2
|
||||
input-object-min-width: 10
|
||||
input-object-min-height: 10
|
||||
model-color-format: 1
|
||||
# 基于模型的编号 0此处为车辆检测模型
|
||||
operate-on-gie-id: 1
|
||||
# 基于模型类别的编号 0此处为车辆类别
|
||||
operate-on-class-ids: 0
|
||||
# 分类器异步模式启动与否
|
||||
classifier-async-mode: 1
|
||||
# 分类器阈值，高于该值则认为分类成功
|
||||
classifier-threshold: 0.51
|
||||
# 基于上游ROI进行分类
|
||||
process-mode: 2
|
||||
#scaling-filter: 0
|
||||
Reference in New Issue
Block a user