README.md

metadata

title: Stashtag Onnx
emoji: 🦀
colorFrom: pink
colorTo: green
sdk: gradio
sdk_version: 5.44.1
app_file: app.py
pinned: false

Stashtag - Video Frame Tagging (ONNX)

🚀 Fast, lightweight ML-powered video frame tagging and marker detection

Features

• ONNX Runtime - Fast inference without PyTorch dependencies
• Modern Gradio 5 - Clean, responsive web interface
• Modular Architecture - Well-organized, maintainable code
• 28 Classes - Comprehensive video frame classification

Quick Start

# Install dependencies
uv sync

# Run the application
uv run python app.py

Usage

1. Tag Mode: Upload sprite image and VTT file to get frame-level tags
2. Marker Mode: Get timestamped markers with confidence scores

Architecture

• stashtag/models/onnx_predictor.py - ONNX-based ML inference
• stashtag/utils/vtt_parser.py - VTT file parsing utilities
• stashtag_model.onnx - Exported model (335MB)
• model_vocab.json - Class mappings

Dependencies

• onnxruntime - ML inference
• gradio>=5.0 - Web interface
• pillow - Image processing
• numpy - Array operations

Clean, minimal dependencies compared to the original PyTorch/FastAI stack.

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference