Baseline

Python Modules

recipes.cad_icassp_2024.baseline.enhance

Run the dummy enhancement.

recipes.cad_icassp_2024.baseline.evaluate

Evaluate the enhanced signals using the HAAQI metric.

recipes.cad_icassp_2024.baseline.merge_batches_results

Join batches scores into a single file.

Configuration Files

config.yaml

 1path:
 2  root: ???
 3  metadata_dir: ${path.root}/metadata
 4  music_dir: ${path.root}/audio/at_mic_music
 5  gains_file: ${path.metadata_dir}/gains.json
 6  head_positions_file: ${path.metadata_dir}/head_positions.json
 7  listeners_file: ${path.metadata_dir}/listeners.train.json
 8  music_file: ${path.metadata_dir}/at_mic_music.train.json
 9  scenes_file: ${path.metadata_dir}/scenes.train.json
10  scene_listeners_file: ${path.metadata_dir}/scene_listeners.train.json
11  exp_folder: ./exp # folder to store enhanced signals and final results
12
13sample_rate: 44100
14remix_sample_rate: 32000
15HAAQI_sample_rate: 24000
16
17nalr:
18  nfir: 220
19  sample_rate: ${sample_rate}
20
21apply_compressor: False
22compressor:
23  threshold: 0.35
24  attenuation: 0.1
25  attack: 50
26  release: 1000
27  rms_buffer_size: 0.064
28
29soft_clip: False
30
31separator:
32  model: demucs   # demucs or openunmix
33  device: ~
34
35evaluate:
36  set_random_seed: True
37  small_test: False
38  batch_size: 1  # Number of batches
39  batch: 0       # Batch number to evaluate
40
41# hydra config
42hydra:
43  run:
44    dir: ${path.exp_folder}
45  job:
46    chdir: True