config.json

{
  "model_type": "humanoid-vision",
  "image_size": 224,
  "hidden_size": 384,
  "num_layers": 6,
  "num_attention_heads": 6,
  "description": "Vision-based perception model for humanoid robots"
}