走向实用化之v3.0

README.md

@@ -1,4 +1,3 @@
-
 ![enter image description here](https://img-blog.csdnimg.cn/7007a6ec9d584018bdf289bd8987c45d.png?x-oss-process=image/watermark,type_ZHJvaWRzYW5zZmFsbGJhY2s,shadow_50,text_Q1NETiBA6YOt5bu65rSL,size_20,color_FFFFFF,t_70,g_se,x_16)
 # [中文] | [[English]](https://github.com/guojianyang/cv-detect-robot/blob/main/README_EN.md)
 # CDR(cv-detect-robot)项目介绍-------（工业级视觉算法侧端部署）
@@ -10,6 +9,19 @@
 >  （ **备注(３)**）:随着本人及团队《中山大学－MICRO_lab实验室》的学习成长，该项目会不定期进行维护和更新，由于能力有限，项目中存在错误和不足之处望各位批评指正或在`issue`中留言。
 
 >  （ **备注(4)**）:为方便大家学习交流，已建立**CDR(cv-detect-robot)项目**交流微信群，请添加群负责人`小郭`微信号`17370042325`,以方便拉您进群。
+***
+***
+# CDR~走向实用化之-v3.0版本更新内容如下🔥🔥🔥🔥🔥：
+- 增加子项目（六）和（七），分别为yolox的python接口和cpp接口。
+- 在子项目(五)中，将摄像头检测和视频文件检测集成到一个python程序中，具体操作参考同级文件夹下README.md。
+- 解决子项目(一)(二)(三)(四)中实时摄像头检测屏幕画面全覆盖问题。
+- 解决频繁出现的`mmap err:Bad file descriptor`错误。
+- 对yolov5和yolox检测模型后都级联了DCF目标跟踪器。
+- 对于所有子项目都可进行指定目标类别检测跟踪。
+- 发布了Nano板和NX板在运行CDR项目时需要注意事项的README.md文档。
+- 发布如何生成engine文件的README.md文档。
+- 解决子项目(二)yolov5-deepstream-python中ros节点读取数据一直显示24个恒定目标数据的问题
+
 ***
 ***
 # CDR子项目(一)（yolov5-ros-deepstream）
@@ -50,5 +62,24 @@
 
 > 最终视频检测效果请进入[resnet10-ros-deepstream检测](https://www.bilibili.com/video/BV1Xg411w78P/)
 
+# CDR子项目(六)（yolox-deepstream-python）
+- yolox-deepstream-python 子项目简介
+> 该项目是将yolox视觉检测算法与神经网络加速引擎tensorRT结合，本子项目采另一种引擎文件生成方法，通过onnx转到engine，此方法更具灵活性，也越来越稳定，符合行业主流发展趋势，在英伟达的deepstream框架下运行，在同一硬件平台上的任意软件目录中，建立一个读取物理内存的`client.py`脚本文件（里面只包含一个读取内存的代码段），将指定好的物理内存中的数据读取出来，在读取成功的前提下，可将该代码段插入到任意需要目标检测数据的python项目中，从而使该python项目能顺利获取目标检测数据。
+
+> 详细教程请进入[yolox-deepstream-python](https://github.com/guojianyang/cv-detect-robot/tree/main/yolox-ros-deepstream)
+
+> 最终视频检测效果请进入[yolox-deepstream-python检测](https://www.bilibili.com/video/BV1k34y1o7Ck/)
+
+# CDR子项目(七)（yolox-deepstream-cpp）
+- yolox-deepstream-cpp 子项目简介
+> 该项目是将yolox视觉检测算法与神经网络加速引擎tensorRT结合，本子项目采另一种引擎文件生成方法，通过onnx转到engine，此方法更具灵活性，也越来越稳定，符合行业主流发展趋势，在英伟达的deepstream框架下运行，在同一硬件平台上的任意软件目录中，建立一个读取物理内存的`yolox_tensor.cpp`文件（里面只包含一个读取内存的代码段），编译后可将指定好的物理内存中的数据读取出来，在读取成功的前提下，可将该代码段插入到任意需要目标检测数据的C++项目中，从而使该C++项目能顺利获取目标检测数据。
+
+> 详细教程请进入[yolox-deepstream-cpp](https://github.com/guojianyang/cv-detect-robot/tree/main/yolox-ros-deepstream)
+
+> 最终视频检测效果请进入[yolox-deepstream-cpp检测](https://www.bilibili.com/video/BV1k34y1o7Ck/)
+
 # [CDR项目常见问题及其解决方案(Common problems and solutions)](https://github.com/guojianyang/cv-detect-robot/wiki/CDR%E9%A1%B9%E7%9B%AE%E5%B8%B8%E8%A7%81%E9%97%AE%E9%A2%98%E5%8F%8A%E5%85%B6%E8%A7%A3%E5%86%B3%E6%96%B9%E6%A1%88(Common-problems-and-solutions))
+# [Jetson Nano和 NX在运行CDR项目时注意事项](https://github.com/guojianyang/cv-detect-robot/wiki/Jetson-Nano%E5%92%8C-NX%E5%9C%A8%E8%BF%90%E8%A1%8CCDR%E9%A1%B9%E7%9B%AE%E6%97%B6%E6%B3%A8%E6%84%8F%E4%BA%8B%E9%A1%B9)
+# [wts文件生成engine文件的方法](https://github.com/guojianyang/cv-detect-robot/wiki/wts%E6%96%87%E4%BB%B6%E7%94%9F%E6%88%90engine%E6%96%87%E4%BB%B6%E7%9A%84%E6%96%B9%E6%B3%95)
+
 

resnet10-ros-deepstream/deepstream_python_apps/apps/deepstream-test7/README.md

@@ -1,3 +1,13 @@
-# test7 建立只能播放的实例，并能实时打印出目标检测数据(坐标、目标跟踪ID号和类别)
+# test7 建立只能播放的实例，并能实时打印出目标检测数据(坐标、置信度和类别)
 
+python3命令后面第一个参数“1”代表视频文件输入模式，“2”代表实时摄像头输入模式
 
+- video_file:
+python3 deepstream-test_7_usb_file.py 1 /opt/nvidia/deepstream/deepstream-5.1/samples/streams/sample_720p.h264
+（备注）：jetson NX 中，以上输入可能会报错，此时需将183行`source_file.set_property('location', args[2])`中的args[2]直接改为视频文件的绝对路径
+
+
+- real_video:
+python3 deepstream-test_7_usb_file.py 2 /dev/video0
+
+

resnet10-ros-deepstream/deepstream_python_apps/apps/deepstream-test7/deepstream-test_7.py → resnet10-ros-deepstream/deepstream_python_apps/apps/deepstream-test7/deepstream-test_7_usb_file.py

@@ -17,6 +17,7 @@
 fps_streams = {}
 frame_count = {}
 saved_count = {}
+Detect_Mode = 0 # 默认值为"0", "1"为 file文件读取模式， "2"为usb实时视频检测模式
 bounding_bboxes =[]
 PGIE_CLASS_ID_VEHICLE = 0
 PGIE_CLASS_ID_BICYCLE = 1
@@ -75,7 +76,7 @@ def osd_sink_pad_buffer_probe(pad, info, u_data):
  bounding_bboxes.append(int(width)) 
  bounding_bboxes.append(int(height))
  bounding_bboxes.append(int(object_id)) 
- obj_meta.rect_params.border_color.set(0.0, 0.0, 0.0, 1.0)
+ obj_meta.rect_params.border_color.set(1.0, 0.0, 0.0, 0.0) #--------red ,green, blue, black
  try:
  l_obj = l_obj.next
 
@@ -133,15 +134,32 @@ def osd_sink_pad_buffer_probe(pad, info, u_data):
  return Gst.PadProbeReturn.OK
 
 
-def main():
+def main(args):
+ if len(args)!=3:
+ sys.stderr.write(" \n\n\n对不起，您还未输入视频源！！！\n\n\n")
+ sys.exit(1)
+ else:
+ if args[1]=="1":
+ Detect_Mode = 1
+ print("您已进入视频文件检测模式！！！") 
+ if args[1]=="2":
+ Detect_Mode = 2
+ print("您已进入实时视频检测模式！！！")
  GObject.threads_init()
  Gst.init(None)
-
  print("Creating Pileline \n")
  pipeline = Gst.Pipeline()
- source = Gst.ElementFactory.make("filesrc", "file-source")
- h264parser = Gst.ElementFactory.make("h264parse", "h264-parper") # h264的编解码
- decoder = Gst.ElementFactory.make("nvv4l2decoder"," nvv4l2-decoder") # h264的编解码
+ if Detect_Mode ==1:
+ source_file = Gst.ElementFactory.make("filesrc", "file-source")
+ h264parser = Gst.ElementFactory.make("h264parse", "h264-parper") # h264的编解码
+ decoder = Gst.ElementFactory.make("nvv4l2decoder"," nvv4l2-decoder") # h264的编解码
+ if Detect_Mode == 2:
+ source_usb = Gst.ElementFactory.make("v4l2src", "usb-cam-source")
+ caps_v4l2src = Gst.ElementFactory.make("capsfilter", "v4l2src_caps")
+ vidconvsrc = Gst.ElementFactory.make("videoconvert", "convertor_src1")
+ nvvidconvsrc = Gst.ElementFactory.make("nvvideoconvert", "convertor_src2")
+ caps_vidconvsrc = Gst.ElementFactory.make("capsfilter", "nvmm_caps")
+
  streammux = Gst.ElementFactory.make("nvstreammux", "Stream-muxer") # autobatch 自动批量处理
  pgie = Gst.ElementFactory.make("nvinfer", "primary-inference") #第一级网络（主要的推理引擎engine）
 
@@ -153,22 +171,29 @@ def main():
  if not sink_real:
  sys.stderr.write(" Unable to create egl sink \n")
 
+
  tracker = Gst.ElementFactory.make("nvtracker", "tracker")
  if not tracker:
  sys.stderr.write(" Unable to create tracker \n")
 
  sgie1= Gst.ElementFactory.make("nvinfer", "secondary1-inference") #第二级网络
  nvvidconv = Gst.ElementFactory.make("nvvideoconvert", "convertor") #转换工具(因为输出的东西要放到画面上)
  nvosd = Gst.ElementFactory.make("nvdsosd"," onscreendisplay") # 显示display工具
+ if Detect_Mode==1:
+ source_file.set_property('location', args[2]) #输入本地文件视频源
+ if Detect_Mode==2:
+ source_usb.set_property('device', args[2]) #输入usb实时视频源
+ caps_v4l2src.set_property('caps', Gst.Caps.from_string("video/x-raw, framerate=30/1"))
+ caps_vidconvsrc.set_property('caps', Gst.Caps.from_string("video/x-raw(memory:NVMM)"))
 
- source.set_property('location', "/opt/nvidia/deepstream/deepstream-5.1/samples/streams/sample_720p.h264") #输入视频源
  streammux.set_property('width', 1920) #设置输入源的宽
- streammux.set_property("height", 1080) #设置输入源的高
- streammux.set_property("batch-size", 1)
+ streammux.set_property('height', 1080) #设置输入源的高
+ streammux.set_property('batch-size', 1)
  streammux.set_property("batched-push-timeout", 4000000)
  pgie.set_property('config-file-path',"dstest7_pgie_config.txt")
  sgie1.set_property('config-file-path',"dstest7_sgie1_config.txt")
  #pgie.set_property('config-file-path', "dstest3_pgie_config.txt")
+ sink_real.set_property('sync', False) #决定实时视频是否同步显示
 
  config = configparser.ConfigParser()
  config.read('dstest7_tracker_config.txt')
@@ -195,25 +220,49 @@ def main():
  tracker.set_property('enable_batch_process', tracker_enable_batch_process)
 
  print("Adding elements to Pipeline \n")
- pipeline.add(source )
- pipeline.add(h264parser)
- pipeline.add(decoder)
+ if Detect_Mode==1:
+ pipeline.add(source_file )
+ pipeline.add(h264parser)
+ pipeline.add(decoder)
+ if Detect_Mode==2:
+ pipeline.add(source_usb)
+ pipeline.add(caps_v4l2src)
+ pipeline.add(vidconvsrc)
+ pipeline.add(nvvidconvsrc)
+ pipeline.add(caps_vidconvsrc)
  pipeline.add(streammux)
  pipeline.add(pgie)
+
  pipeline.add(tracker)
+
  pipeline.add(sgie1)
  pipeline.add(nvvidconv)
  pipeline.add(nvosd) 
  if is_aarch64():
  pipeline.add(transform)
  pipeline.add(sink_real)
 
- source.link(h264parser)
- h264parser.link(decoder)
- sinkpad = streammux.get_request_pad("sink_0")
- srcpad = decoder.get_static_pad("src")
- srcpad.link(sinkpad)
+ if Detect_Mode==1:
+ source_file.link(h264parser)
+ h264parser.link(decoder)
+ sinkpad_file = streammux.get_request_pad("sink_0")
+ srcpad_file = decoder.get_static_pad("src")
+ srcpad_file.link(sinkpad_file)
+ if Detect_Mode==2:
+ source_usb.link(caps_v4l2src)
+ caps_v4l2src.link(vidconvsrc)
+ vidconvsrc.link(nvvidconvsrc)
+ nvvidconvsrc.link(caps_vidconvsrc)
+ sinkpad_usb = streammux.get_request_pad("sink_0")
+ if not sinkpad_usb:
+ sys.stderr.write(" Unable to get the sink pad of streammux \n")
+ srcpad_usb = caps_vidconvsrc.get_static_pad("src")
+ if not srcpad_usb:
+ sys.stderr.write(" Unable to get source pad of caps_vidconvsrc \n")
+ srcpad_usb.link(sinkpad_usb)
  streammux.link(pgie)
+
+
  pgie.link(tracker)
  tracker.link(sgie1)
  sgie1.link(nvvidconv)
@@ -223,6 +272,7 @@ def main():
  transform.link(sink_real)
  else:
  nvosd.link(sink_real)
+
  loop = GObject.MainLoop()
  bus = pipeline.get_bus()
  bus.add_signal_watch()
@@ -245,4 +295,8 @@ def main():
  pipeline.set_state(Gst.State.NULL)
 
 if __name__ == "__main__":
- sys.exit(main())
+ sys.exit(main(sys.argv))
+
+
+
+

resnet10-ros-deepstream/deepstream_python_apps/apps/deepstream-test7/dstest7_pgie_config.txt

@@ -60,26 +60,24 @@
 [property]
 gpu-id=0
 net-scale-factor=0.0039215697906911373
-
 model-file=../../../../samples/models/Primary_Detector/resnet10.caffemodel
 proto-file=../../../../samples/models/Primary_Detector/resnet10.prototxt
-model-engine-file=../../../../samples/models/Primary_Detector/resnet10.caffemodel_b1_gpu0_fp16.engine
+model-engine-file=../../../../samples/models/Primary_Detector/resnet10.caffemodel_b1_gpu0_int8.engine
 labelfile-path=../../../../samples/models/Primary_Detector/labels.txt
-
+#int8-calib-file=../../../../samples/models/Primary_Detector/cal_trt.bin
 force-implicit-batch-dim=1
 batch-size=1
-network-mode=2
-process-mode=1
-model-color-format=0
+network-mode=1
 num-detected-classes=4
-interval=5
+interval=1
 gie-unique-id=1
 output-blob-names=conv2d_bbox;conv2d_cov/Sigmoid
 
 #scaling-filter=0
 #scaling-compute-hw=0
 
 [class-attrs-all]
-pre-cluster-threshold=0.2
+pre-cluster-threshold=0.4
 eps=0.2
 group-threshold=1
+

resnet10-ros-deepstream/deepstream_python_apps/apps/deepstream-test7/dstest7_sgie1_config.txt

@@ -62,34 +62,24 @@ gpu-id=0
 net-scale-factor=1
 model-file=../../../../samples/models/Secondary_CarColor/resnet18.caffemodel
 proto-file=../../../../samples/models/Secondary_CarColor/resnet18.prototxt
-model-engine-file=../../../../samples/models/Secondary_CarColor/resnet18.caffemodel_b16_gpu0_fp32.engine
+model-engine-file=../../../../samples/models/Secondary_CarColor/resnet18.caffemodel_b16_gpu0_int8.engine
 mean-file=../../../../samples/models/Secondary_CarColor/mean.ppm
 labelfile-path=../../../../samples/models/Secondary_CarColor/labels.txt
-int8-calib-file=../../../../samples/models/Secondary_CarColor/cal_trt.bin
-
+#int8-calib-file=../../../../samples/models/Secondary_CarColor/cal_trt.bin
 force-implicit-batch-dim=1
 batch-size=16
 # 0=FP32 and 1=INT8 mode
-network-mode=0
+network-mode=1
 input-object-min-width=64
 input-object-min-height=64
 process-mode=2
 model-color-format=1
 gpu-id=0
-
-#本模型的id号
 gie-unique-id=2
-
-#接收哪个推理模型输出的数据
 operate-on-gie-id=1
-
-#所检测的类别id（PGIE_CLASS_ID_VEHICLE = 0）
 operate-on-class-ids=0
 is-classifier=1
-
-# 输出层
 output-blob-names=predictions/Softmax
-
 classifier-async-mode=1
 classifier-threshold=0.51
 process-mode=2

resnet10-ros-deepstream/deepstream_python_apps/apps/deepstream-test7/dstest7_tracker_config.txt

@@ -28,11 +28,13 @@
 # ll-config-file: required for NvDCF, optional for KLT and IOU
 #
 [tracker]
-tracker-width=640
-tracker-height=384
+#tracker-width=640
+#tracker-height=384
+tracker-width=960
+tracker-height=544
 gpu-id=0
-ll-lib-file=/opt/nvidia/deepstream/deepstream-5.1/lib/libnvds_mot_klt.so
-#ll-lib-file=/opt/nvidia/deepstream/deepstream-5.1/lib/libnvds_nvdcf.so
+#ll-lib-file=/opt/nvidia/deepstream/deepstream-5.1/lib/libnvds_mot_klt.so
+ll-lib-file=/opt/nvidia/deepstream/deepstream-5.1/lib/libnvds_nvdcf.so
 ll-config-file=tracker_config.yml
 #enable-past-frame=1
 enable-batch-process=1