🖼️ Vision-GPT Image Captioning

Upload an image and generate a descriptive caption using Vision-GPT (GPT-2 + ViT + Cross-Attention trained on Flickr8k)
10 100
0.1 2
0.1 1
0 100

📊 Provide Feedback

Help improve the model! Enter the correct caption and we'll calculate similarity.

📈 Feedback Results

Model Info

  • Architecture: GPT-2 (124M) + ViT-B/16 + Cross-Attention
  • Training: Flickr8k dataset (8k images)

Similarity Scoring

  • ≥80%: 🌟 High Reward (+10)
  • ≥60%: 👍 Good Reward (+5)
  • ≥50%: 🤔 Low Reward (+2)
  • <40%: 😞 Penalty (-5)