Frequently Asked Questions (FAQ)

Common questions and troubleshooting tips.

---

🚀 Getting Started

Q: What's the minimum hardware requirement?

A: For basic usage:

• CPU: Any modern processor
• RAM: 4GB minimum, 8GB recommended
• Storage: 2GB free space

For GPU acceleration:

• NVIDIA: GTX 1060 6GB or better
• Apple Silicon: M1 or newer

---

🐛 Installation Issues

Q: CUDA is available but not detected

A: Verify PyTorch CUDA:

python -c "import torch; print(torch.cuda.is_available())"

If False, reinstall with correct CUDA version:

pip install torch torchvision --index-url https://download.pytorch.org/whl/cu118

---

⚡ Performance

Q: How to improve inference speed?

A: Try these optimizations:

1. Use smaller model:

   MODEL_NAME=yolov8n.pt

2. Reduce resolution:

   canvas.toBlob(sendFrame, 'image/jpeg', 0.7)

3. Enable GPU:

   DEVICE=cuda:0

Q: Out of memory errors

A: Reduce memory usage:

MODEL_NAME=yolov8n.pt
MAX_CONCURRENCY=1

---

🎯 Model Usage

Q: How to detect custom objects?

A: Use open-vocabulary detection:

curl -X POST "http://localhost:8000/infer" \
  -F "model=google/owlvit-base-patch32" \
  -F "file=@image.jpg" \
  -F "text_queries=your custom object"

---

🔗 Still Need Help?

• Documentation
• GitHub Issues