This repository hosts the official implementation of Efficient Feature Aggregation and Scale-Aware Regression for Monocular 3-D Object Detection.

The official results in the paper on KITTI Val Set:

This repo results on KITTI Val Set:

1. Clone this project and create a conda environment:

   git clone https://github.com/WYFDUT/MonoASRH.git
   cd MonoASRH
   
   conda create -n monoasrh python=3.9
   conda activate monoasrh

2. Install pytorch and torchvision matching your CUDA version:

   # For example, We adopt torch 1.11.0+cu113
   pip install torch==1.11.0+cu113 torchvision==0.12.0+cu113 torchaudio==0.11.0 --extra-index-url https://download.pytorch.org/whl/cu113

3. Install requirements:

   pip install -r requirements.txt

4. Download KITTI datasets and prepare the directory structure as:

   │MonoASRH/
   ├──...
   │data/kitti/
   ├──ImageSets/
   ├──training/
   │   ├──image_2
   │   ├──label_2
   │   ├──calib
   ├──testing/
   │   ├──image_2
   │   ├──calib

   You can also change the data path at "dataset/root_dir" in lib/kitti.yaml.

You can modify the settings of models and training in lib/kitti.yaml:

python tools/train_val.py

python tools/train_val.py -e

The best checkpoint will be evaluated as default. You can change it at "tester/resume_model" in lib/kitti.yaml:

python tools/train_val.py -t

If you use this code in your research, please cite:

@ARTICLE{11395320,
  author={Wang, Yifan and Yang, Xiaochen and Pu, Fanqi and Liao, Qingmin and Yang, Wenming},
  journal={IEEE Transactions on Intelligent Transportation Systems}, 
  title={Efficient Feature Aggregation and Scale-Aware Regression for Monocular 3-D Object Detection}, 
  year={2026},
  volume={},
  number={},
  pages={1-14},
  keywords={3D object detection;monocular;scale-aware;scene understanding;autonomous driving},
  doi={10.1109/TITS.2026.3659175}}

This repo benefits from the excellent work GUPNet and MonoLSS