🔍 ZeroDINO [MMM'26]🎉

ZeroDINO: Entropy-Driven Granularity-Aware Semantic Fusion for Zero-Shot Learning

---

🧠 Model Architecture

🚀 Quick Start

Before you begin, please make sure you have downloaded the following datasets:

• 🐦 CUB-200-2011
• 🌞 SUN Attribute
• 🐘 AWA2

✅ 1. Install Dependencies

Install all required Python packages:

pip install -r requirements.txt

✅ 2. Train or Evaluate

Train on a specific dataset:

bash train.sh CUB      # or SUN / AWA2

Evaluate using pretrained weights:

bash test.sh CUB      # or SUN / AWA2

📈 Results

Performance of our released models on three benchmark datasets under two evaluation protocols: Conventional Zero-Shot Learning (CZSL) and Generalized Zero-Shot Learning (GZSL).

Dataset	Acc (CZSL)	Unseen (GZSL)	Seen (GZSL)	Harmonic Mean (H)
CUB	86.6	78.3	82.7	80.4
SUN	79.3	57.1	52.0	54.4
AWA2	73.9	66.1	86.9	75.1