mhussainahmad/averformer-ravdess-v4

AVERFormer-v4 (RAVDESS)

Multimodal Audio-Visual-Text Emotion Recognition transformer trained on RAVDESS. Code: https://github.com/mhussainahmad/AVERFormer

Reported numbers

• Best single-seed val wF1: 65.9090909090909 (seed 42, epoch 19)
• Best ensemble wF1: 0.7887 (top-3 ensemble)

Classes (7)

['neutral', 'happy', 'sad', 'angry', 'fearful', 'disgust', 'surprised']

Protocol: 7-class RAVDESS (calm dropped, consistent with cross-corpus AVR convention).

Architecture

• Audio: microsoft/wavlm-large (16 kHz mono waveform)
• Video: MCG-NJU/videomae-large (16 frames @ 224x224 RGB)
• Text: microsoft/deberta-v3-large (speaker-aware ctx encoder)
• Fusion: 2-layer cross-modal transformer, dim=512, 8 heads
• Heads: face / voice / text / joint (all share class count)

Loading

import json, torch
from huggingface_hub import hf_hub_download
from models.averformer_v4 import AVERFormerV4

cfg = json.load(open(hf_hub_download(repo_id="mhussainahmad/averformer-ravdess-v4", filename="config.json")))
ckpt = hf_hub_download(repo_id="mhussainahmad/averformer-ravdess-v4", filename="pytorch_model.pth")

model = AVERFormerV4(
    audio_backbone=cfg["audio_backbone"],
    video_backbone=cfg["video_backbone"],
    text_backbone=cfg["text_backbone"],
    num_classes=cfg["num_classes"],
    fusion_layers=cfg["fusion_layers"],
    lora_r=cfg["lora_r"],
    use_text=True,
)
state = torch.load(ckpt, map_location="cpu", weights_only=False)
model.load_state_dict(state["model"], strict=False)
model.eval()

Live inference

python live_emotion_v4.py --repo_id mhussainahmad/averformer-ravdess-v4

See LIVE_INFERENCE_README.md in the GitHub repo for full setup.