The KITTI Vision Benchmark Suite

Most one-stage 3D object detectors complete
different tasks with the same extracted features.
Nevertheless, it is impossible to project features
into a common space that is adaptive for all the
tasks. We present a novel task-aware deformable
prediction (TADP) method for single-stage 3D
object detection to solve this problem. Firstly, a
triple feature refinement aggregation module is
designed to extract three-level features
adaptively. Additionally, we design the multi-
scale feature aggregation block to fuse multi-
scale features in a scale-aware manner. Finally,
the prediction of each task is deformed with the
designed plug-and-play task-aware deformation head
to make the detector head predicts more accurate
results. The perception stack in the head can
percept the emphasis of each task and provide
better interaction among the tasks. The
deformation map can learn the predicted
deformation of tasks through four different
deformation modules.

\epoch=60
\GPU=RTX3090
\LR=0.003
\decay factor=0.4
\Batch size=4