metadata
license: mit
tags:
- robotics
- humanoid
- action-recognition
- deep-learning
- computer-vision
pipeline_tag: image-classification
Humanoid Action Classifier v1
This model is designed for humanoid robot action recognition. It classifies basic humanoid movements from image input.
Supported Actions
- walk
- run
- sit
- stand
- wave
- pick_object
- turn_left
- turn_right
Model Details
- Architecture: CNN (ResNet18 backbone)
- Framework: PyTorch
- Input Size: 224x224 RGB
- Output: 8 action classes
Training Info
Trained on a synthetic humanoid action dataset. Optimized for robotics simulation environments.
Usage
from transformers import AutoModelForImageClassification
from PIL import Image
import torch
model = AutoModelForImageClassification.from_pretrained("your-username/humanoid-action-classifier-v1")
image = Image.open("test.jpg")
inputs = processor(images=image, return_tensors="pt")
with torch.no_grad():
outputs = model(**inputs)
predicted_class = outputs.logits.argmax(-1)