EffiDec3D

Official Pytorch implementation of the paper EffiDec3D: An Optimized Decoder for High-Performance and Efficient 3D Medical Image Segmentation published in CVPR 2025. code video
Md Mostafijur Rahman, Radu Marculescu

The University of Texas at Austin

🔍 Check out our NeurIPS 2025 paper! LoMix

🔍 Check out our MICCAI 2025 paper! EfficientMedNeXt

🔍 Check out our ICCVW 2025 paper! MK-UNet

🔍 Check out our CVPR 2024 paper! EMCAD

🔍 Check out our CVPRW 2024 paper! PP-SAM

🔍 Check out our WACV 2024 paper! G-CASCADE

🔍 Check out our MIDL 2023 paper! MERIT

🔍 Check out our WACV 2023 paper! CASCADE

Update

🚀 July 20, 2025: Code released for EffiDec3D optimized networks!!!

Usage:

Recommended environment:

Please run the following commands.

conda create -n effidec3denv python=3.8
conda activate effidec3denv

Training:

cd into EffiDec3D
python main_finetune_BTCV_TU.py --root </data/btcv_trns/> --output output_folder/run1 --dataset BTCV13 --img_size 96 96 96 --n_channels 1 --network 3DUXNET_EffiDec3D --channels 48 96 192 384 --n_decoder_channels 48 --ds False --mode train --pretrain False --batch_size 1 --crop_sample 4 --lr 0.001 --optim AdamW --max_iter 45000  --eval_step 250 --val_batch 1 --gpu 0 --cache_rate 1.0 --num_workers 4 --overlap 0.7 #> output_folder/BTCV13_3DUXNET_EffiDec3D_loss_dsFalse_1out_96x96x96_lr1e3_itr45000_overlap070_run1.txt

Testing:

cd into EffiDec3D 

Acknowledgement

We are very grateful for these excellent works monai, 3D UX-Net, SwinUNETR, and MedNeXt, which have provided the basis for our framework.

Citations

@inproceedings{rahman2025effidec3d,
  title={EffiDec3D: An Optimized Decoder for High-Performance and Efficient 3D Medical Image Segmentation},
  author={Rahman, Md Mostafijur and Marculescu, Radu},
  booktitle={Proceedings of the Computer Vision and Pattern Recognition Conference},
  pages={10435--10444},
  year={2025}
}