unigaze 0.1.1

UniGaze: lightweight loader that pulls pretrained weights from Hugging Face

UniGaze Easy Loader

A tiny, dependency-light Python package to load UniGaze pretrained models from Hugging Face.

📦 Installation

Install a matching PyTorch first.

** CUDA 12.8 example**

pip install torch==2.7.0 torchvision==0.22.0 --index-url https://download.pytorch.org/whl/cu128
pip install unigaze

🚀 Quick Start

import torch
model = unigaze.load("unigaze_h14_joint", device="cuda")   # downloads weights from HF on first use
# Input: normalized batch (B, 3, 224, 224)
image_normalized_batch = torch.ones((10, 3, 224, 224), device="cuda")
# Output: {'pred_gaze': (B, 2)} with (pitch, yaw)
pred_gaze = model(image_normalized_batch)['pred_gaze']
print(pred_gaze.shape)  # torch.Size([10, 2])