Skip to content

Latest commit

 

History

History
29 lines (22 loc) · 1.13 KB

README.md

File metadata and controls

29 lines (22 loc) · 1.13 KB

Segment Anything for Visual Bird Sound Denoising

Implementation of "Segment Anything for Visual Bird Sound Denoising".

Installation

pip install -r requirements.txt

Model

We use the ViT-B SAM model to initialize the BudSAM model parameters.

We provide access to our trained BudSAM model.

Run

nohup python budsam_train_multi.py \
    -task_name BudSAM-test-5GPUs \
    -work_dir ./work_dir \
    -batch_size 2 \
    -num_workers 5 \
    --world_size 5 \
    --bucket_cap_mb 25 \
    --grad_acc_steps 1 \
    --node_rank 0 \
    --init_method tcp://localhost:12344 > ./logs/run.log 2>&1 &

Acknowledgements

  • We thank the BirdSoundsDenoising dataset contributors for providing the dataset to the community.
  • We thank for the source code of Segment Anything publicly available.
  • We thank for the source code of Segment anything in medical images publicly available.