ultralytics 8.3.219

v8.3.219 - `ultralytics 8.3.219` FastSAM masks `.float()` upsampling fix (#22460)

on Python PyPI

🌟 Summary

FastSAM gets a reliable mask-resizing fix and a cleaner, built-in CLIP integration for prompt-based segmentation—improving accuracy, stability, and ease of use. 🚀🖼️

📊 Key Changes

• 🧩 FastSAM mask upsampling fix
  • Masks are now converted to float before upsampling and then thresholded back to byte: ensures accurate, error-free resizing during prompts.
• 🤖 Built-in CLIP integration
  • Replaced external CLIP dependency with the internal ultralytics.nn.text_model.CLIP, standardizing image/text feature extraction and similarity scoring.
• 🔧 Version bump
  • Updated package version to 8.3.219.

Reference: See the current PR FastSAM masks .float() upsampling fix (PR #22460) by @Y-T-G.

🎯 Purpose & Impact

• ✅ More accurate segmentation prompts
  • Prevents artifacts and type-related errors during mask resizing, yielding cleaner, more reliable masks in FastSAM workflows.
• 🛡️ Improved robustness and compatibility
  • Reduces issues across different PyTorch versions and image sizes by using correct tensor types and thresholding.
• 🧱 Simpler, more reliable text prompts
  • Built-in CLIP removes ad-hoc installs and runtime imports, improving setup speed, offline usability, and consistency across environments.
• 💼 No action needed for most users
  • Existing FastSAM usage benefits automatically; prompt-based segmentation should be more stable out of the box.

What's Changed

• ultralytics 8.3.219 FastSAM masks .float() upsampling fix by @Y-T-G in #22460

Full Changelog: v8.3.218...v8.3.219