🦉 NoctOWLv2: Fine-Grained Open-Vocabulary Object Detection

NoctOWL (Not only coarse-text OWL) extends OWL-ViT and OWLv2 for Fine-Grained Open-Vocabulary Detection (FG-OVD).
It can recognize subtle object differences such as color, texture, and material, while retaining strong coarse-grained detection abilities.

Available Models:

  • 🧩 NoctOWLv2-Base — Smaller and faster.
  • 🧠 NoctOWLv2-Large — More accurate, higher capacity.

📘 Training & evaluation code
0 1
Select Model

Select which model to use for detection
Examples
Input Image Text Queries (comma-separated) Score Threshold