Run with ResNet34 embeddings (baseline)

python run_diarization.py -i audio.wav -o output/ --batch-size 8

Run with ReDimNet embeddings

python run_redimnet_direct_replace_fixed.py -i audio.wav -o output_redimnet/ --redimnet-model b1

Custom Modifications

This repository includes a modified version of diarizen/pipelines/inference.py to support ReDimNet embedding integration.

Changes made:

• Replaced ResNet34 embedding extraction with ReDimNet
• Added projection layer for dimension matching
• Modified clustering input handling

Audio Input → DiariZen Pipeline → VAD + Segmentation → Embedding Extraction (ReDimNet/ResNet34) → VBx Clustering → RTTM Output