v1.5.0

From this version, we adopt semantic versioning for faster and more friendly development experience.

What's Changed

• Adding Microsoft SECURITY.MD by @microsoft-github-policy-service in #2
• nit: fix spelling for requirements.txt by @nmstoker in #17
• Update requirement.txt by @krishna2 in #16
• Updated einops typo in requirements.txt by @redron in #28
• Add torch.inference mode by @aliencaocao in #29
• Add PaddleOCR option by @aliencaocao in #53
• Add icon detect image size option by @aliencaocao in #72
• version 1.5 by @yadong-lu in #94

New Contributors

• @microsoft-github-policy-service made their first contribution in #2
• @nmstoker made their first contribution in #17
• @krishna2 made their first contribution in #16
• @redron made their first contribution in #28
• @aliencaocao made their first contribution in #29
• @yadong-lu made their first contribution in #94

Full Changelog: https://github.com/microsoft/OmniParser/commits/v1.5.0

v.2.0.0

What's new in V2.0.0?

• Larger and cleaner set of icon caption + grounding dataset
• 60% improvement in latency compared to V1 model checkpoints
• Strong performance: 39.6 average accuracy on ScreenSpot Pro
• Your agent only need one tool: OmniTool. Control a Windows 11 VM with OmniParser + your vision model of choice. OmniTool supports out of the box the following large language models - OpenAI (4o/o1/o3-mini), DeepSeek (R1), Qwen (2.5VL) or Anthropic Computer Use.