This is just a simple visualization application for textvqa and textcaps evaluation result built by gradio. Maybe really useful for debugging evaluation error causes.
Features:
- TextVQA: OCR input, image input, question input, and answer prediction and answer groundtruth output
- TextCaps: OCR input, image input, caption prediction and caption groundtruth output
- Random select with index binding
- Select with index slider
- Select split and different ocr prompt subfolder
- two or three different prompt contrast
Usage:
pip install gradio
python visualize_gradio_app.py