can you tell me how to use the object detector details, Thanks you.

[wt2excellent]

No description provided.

[Bradgers]

Sorry for the late reply. You can use a LiDAR-based 3D detector (such as the model in OpenPCDet) and then publish the detection results using ROS. The code is as follows:

from std_msgs.msg import Float64MultiArray
from sensor_msgs.msg import PointCloud2
import ros_numpy

# publisher
detect3d_pub  = rospy.Publisher("/detect3d", Float64MultiArray)

# specify your detector
detector = YOUR_DETECTOR()


def get_boxes(databin):
    result = detector.detect(databin) # [[type, x, y, z, l, w, h, yaw, score], ...,[type, x, y, z, l, w, h, yaw, score]]
    return result


def Callback_pcvelo(msg):
    assert isinstance(msg, PointCloud2)
    # receive pointcloud data
    points = ros_numpy.numpify(msg)
    frame_timestamp = msg.header.stamp.to_sec()
    # get detection results
    result = np.array(get_boxes(points))
    result_1d = np.array(result).reshape(-1)
    data_send = np.array([frame_timestamp] + list(result_1d)) # [timestamp, [type, x, y, z, l, w, h, yaw, score], ...,[type, x, y, z, l, w, h, yaw, score]]
    # publish detection results
    detect3d_pub.publish(Float64MultiArray(data=data_send))
    

if __name__ == "__main__":
    # subscribe to the point cloud topic
    sub_pcvelo = rospy.Subscriber("velodyne_points", PointCloud2, Callback_pcvelo)
    rospy.spin()