multiple cameras

[qwedaq]

Hi , I wanted to know how do I launch detectnet node for multiple camera images at the same time. I have 4 camera image topics with me but how can I run object detection node on all the images at the same time. I am a beginner to this , can you please help me?

Thanks in advance

[r91andersson]

What I've found out to work was to modify the code inside node_detectnet.cpp. In my case I have 3 camera live feed I want to run detection on, so what I did was to add 2 more subscribers (basically just copy paste the first subscriber). ROS_GET_PARAMETER("topic_image_in_camera_1", topic_image_in_camera_1); ROS_GET_PARAMETER("topic_image_in_camera_2", topic_image_in_camera_2); ROS_GET_PARAMETER("topic_image_in_camera_3", topic_image_in_camera_3);

Then,

` /*

• subscribe to image topic */

  auto img_sub_1 = ROS_CREATE_IMAGE_TRANSPORT_SUBSCRIBER(sensor_msgs::Image, topic_image_in_camera_1, 5, img_callback_camera_1);

  auto img_sub_2 = ROS_CREATE_IMAGE_TRANSPORT_SUBSCRIBER(sensor_msgs::Image, topic_image_in_camera_2, 5, img_callback_camera_2);

  auto img_sub_3 = ROS_CREATE_IMAGE_TRANSPORT_SUBSCRIBER(sensor_msgs::Image, topic_image_in_camera_3, 5, img_callback_camera_3);`

They will all use the same model to run the inference on. i.e they all call this: const int numDetections = net->Detect(input_cvt->ImageGPU(), input_cvt->GetWidth(), input_cvt->GetHeight(), &detections, detectNet::OVERLAY_NONE);

However, I'm not sure if there is any guarantee that a specific callback will have the resource to the detection model and be able to run the inference? Maybe you know this @dusty-nv. Will it be a blocking/wait call if the resource (detection model) is already taken by another thread (callback)?

[dusty-nv]

Will it be a blocking/wait call if the resource (detection model) is already taken by another thread (callback)? Yes detectNet::Detect() is a blocking call

[r91andersson]

Ok, so if one thread is using the model and blocking the resource, will the other thread wait until the resource becomes free, or will it continue execute?

[dusty-nv]

Another thread should be able to use it, but I haven't tested that scenario.

From: r91andersson @.> Sent: Monday, May 3, 2021 9:32 AM To: dusty-nv/ros_deep_learning @.> Cc: Dustin Franklin @.>; Mention @.> Subject: Re: [dusty-nv/ros_deep_learning] multiple cameras (#34)

Ok, so if one thread is using the model and blocking the resource, will the other thread wait until the resource becomes free, or will it continue execute?

- You are receiving this because you were mentioned. Reply to this email directly, view it on GitHubhttps://github.com/dusty-nv/ros_deep_learning/issues/34#issuecomment-831261217, or unsubscribehttps://github.com/notifications/unsubscribe-auth/ADVEGK7QH6AOYASFHSUD7Z3TL2QT3ANCNFSM4MKUTM3A.