MonoPlace3D

MonoPlace3D: Learning 3D-Aware Object Placement for 3D Monocular Detection (CVPR 2025) [paper]
Rishubh Parihar, Srinjay Sarkar, Sarthak Vora, Jogendra Kundu, R. Venkatesh Babu.

Setup

Please refer to INSTALL.md for installation and to DATA.md for data preparation.

Train Placement Network

Move to root and train the network with $EXP_NAME:

 cd Placement #MonoPlace3D_ROOT
 CUDA_VISIBLE_DEVICES=$GPUS python scripts/train.py --config=$CONFIG_PATH --experiment_name=$EXP_NAME

Eval Placement Network

To evaluate on the validation set using checkpoint $CHECKPOINT_PATH:

CUDA_VISIBLE_DEVICES=$GPU python3 scripts/eval.py --config=$CONFIG_PATH --gpu=0 --checkpoint_path=$CKPT_PATH --split_to_test=$SPLIT 

Download Blender and ShapNet dataset

wget -c https://download.blender.org/release/Blender2.78/blender-2.78-linux-glibc219-x86_64.tar.bz2

wget -c https://download.blender.org/release/Blender2.78/blender-2.78-linux-glibc219-x86_64.tar.bz2

Render ShapeNet Cars

cd Rendering
blender <PATH TO BLENDER FILE> --background --python KITTI_shapenet_render_cars.py --root=$LABEL_PATH --render_path=$RENDERED_IMAGE_PATH --start_idx 0

Render Car Shadows

cd Rendering
blender <PATH TO BLENDER FILE> --background --python KITTI_shapenet_render_shadows.py --root=$LABEL_PATH --render_path=$RENDERED_IMAGE_PATH --start_idx 0

Place rendered ShapeNet/ControlNet Cars with Shadow in the scene

cd Rendering
python render_cars_with_shadow.py --label_path=$LABEL_PATH --img_path=$IMAGE_2_PATH --render_car_path=$RENDER_CAR_PATH --render_transformed_cars_path=$RENDER_CARS_TRANS_PATH --img_save_path=$IMG_PATH --start_idx 0

Train Object Detection Network

Please refer to the code of MonoDLE and GUPNet.

Citation

If you find our work useful in your research, please consider citing:

@misc{rishubh2025monoplace3D,
      title={MonoPlace3D: Learning 3D-Aware Object Placement for 3D Monocular Detection},
      author={Rishubh Parihar,Srinjay Sarkar,Sarthak Vora,Jogendra Kundu,R. Venkatesh Babu},
      journal={Conference on Computer Vision and Pattern Recognition},      
      year={2025}, 
}

License

This project is released under the MIT License.