camera_streampetr / ml_package_camera_streampetr.param.yaml
xmfcx's picture
feat: add camera_streampetr v1.0 artifacts (from awf.ml.dev.web.auto/perception/models/streampetr/v1)
90f37d6 verified
Raw
History Blame Contribute Delete
863 Bytes
/**:
ros__parameters:
model_params: # These are parameters that are dependent on the onnx file.
class_names: ["CAR", "TRUCK", "BUS", "BICYCLE", "PEDESTRIAN"]
detection_range: [-61.2, -61.2, -10.0, 61.2, 61.2, 10.0] # [-x,-y,-z,x,y,z] from base_link. The model is not trained to predict objects outside this range.
# So, anything outside this range is wrong, and this parameter is used to filter out objects that are outside this range
input_image_height: 480
input_image_width: 640
num_proposals: 5400
rois_number: 5 # ["CAM_FRONT", "CAM_FRONT_LEFT", "CAM_BACK_LEFT", "CAM_FRONT_RIGHT", "CAM_BACK_RIGHT"]
pre_memory_length: 1024 # Input memory length for the temporal model
post_memory_length: 1280 # Output memory length for the temporal model