Learning2HearWhileWalking

This repository provides the G1 and Go1 Transfer-DiT checkpoints used with EgoGraph-mined pseudo-supervision for legged-robot ego-noise separation.

Checkpoints

Robot	File	Best epoch	BAL rank	LoRA rank
G1	checkpoints/G1/best.pt	15	8	16
Go1	checkpoints/Go1/best.pt	18	8	16

Code

The EgoGraph and Transfer-DiT implementations are available at:

https://github.com/blind-review-demo/Learning2HearWhileWalking

SAM-Audio code and base model weights are not included in this repository. Install SAM-Audio and download the base model as described in the code repository.

Usage

git clone https://github.com/blind-review-demo/Learning2HearWhileWalking
cd Learning2HearWhileWalking
bash scripts/setup_sam_audio.sh
huggingface-cli login
python scripts/download_sam_audio_weights.py --model-id facebook/sam-audio-small
python scripts/separate_audio.py \
  --audio path/to/mixed.wav \
  --checkpoint path/to/best.pt \
  --out-dir outputs/demo

Files

Each file is a training checkpoint containing the model state and training metadata. SHA256 checksums are provided in SHA256SUMS.