fast_rcnn.py

What is this?

../../../../_images/fast_rcnn1.gif

Publish an image with object bounding boxes, scores and labels.

CLASSES = ('__background__',
           'aeroplane', 'bicycle', 'bird', 'boat',
           'bottle', 'bus', 'car', 'cat', 'chair',
           'cow', 'diningtable', 'dog', 'horse',
           'motorbike', 'person', 'pottedplant',
           'sheep', 'sofa', 'train', 'tvmonitor')

Subscribing Topic

  • ~input (sensor_msgs/Image)

    Raw image.

  • ~input/rect_array (jsk_recognition_msgs/RectArray)

    Object location proposals.

Publishing Topic

  • ~output (sensor_msgs/Image)

    Image with bounding boxes, scores and labels for objects listed Object Labels.

Parameters

  • ~model (String, required)

    Network model name. (vgg_cnn_m_1024 or vgg16) vgg_cnn_m_1024 is small network and requires ~2GB GPU memory. vgg16 is large network and requires ~5GB GPU memory.

  • ~approximate_sync (Bool, default: False)

    Whether to use approximate for input topics.

  • ~queue_size (Int, default: 10)

    How many messages you allow about the subscriber to keep in the queue. This should be big when there is much difference about delay between two topics.

  • ~slop (Float, default: 0.1)

    How many seconds you allow about the difference of timestamp when you specify ~approximate_sync.

Example

../../../../_images/fast_rcnn_example.jpg
roslaunch jsk_perception sample_fast_rcnn.launch