Genre Identification Deep Checkpoints

Summary

• Version: V16
• Live Space inference: ResNet50 + EfficientNet-B4 + EfficientNet-B3 ensemble on mel-spectrogram heatmaps
• Primary notebook reference: notebooks/dl-21f2000735-notebook-t12026 (12).ipynb
• Feature pipeline: 30s audio -> mel256 spectrogram image -> ResNet50/EfficientNet ensemble
• Tracked public leaderboard score: 0.23380

Labels

• 0 -> blues
• 1 -> classical
• 2 -> country
• 3 -> disco
• 4 -> hiphop
• 5 -> jazz
• 6 -> metal
• 7 -> pop
• 8 -> reggae
• 9 -> rock

Local demo predictions

• demo_audio_pairs/auto_sample/original.wav -> classical
• demo_audio_pairs/auto_sample/messy.wav -> classical

Files

• deep_models/efficientnet_b4_best_2.pth: main CNN checkpoint used by the Space
• deep_models/efficientnet_b3_best_2.pth: secondary CNN checkpoint used by the Space
• deep_models/resnet50_best_2.pth: additional CNN checkpoint used by the Space
• artifacts/rf_baseline_best_public_summary.json: local validation summary

Usage