Skip to content

fix(deps): update dependency ultralytics ( 8.3.179 → 8.3.189 ) - #98

Merged
jazzlyn-bot[bot] merged 1 commit into
mainfrom
renovate/ultralytics-8.x
Aug 30, 2025
Merged

fix(deps): update dependency ultralytics ( 8.3.179 → 8.3.189 )#98
jazzlyn-bot[bot] merged 1 commit into
mainfrom
renovate/ultralytics-8.x

Conversation

@jazzlyn-bot

@jazzlyn-bot jazzlyn-bot Bot commented Aug 18, 2025 •
edited
Loading

Copy link
Copy Markdown
Contributor

This PR contains the following updates:

Package Type Update Change OpenSSF
ultralytics (changelog) project.dependencies patch ==8.3.179 -> ==8.3.189 OpenSSF Scorecard

Warning

Some dependencies could not be looked up. Check the Dependency Dashboard for more information.

Release Notes

ultralytics/ultralytics (ultralytics)

v8.3.189: - ultralytics 8.3.189 3x faster inplace model.fuse() (#​21844)

Compare Source

🌟 Summary

ultralytics 8.3.189 delivers a 3x faster in-place model.fuse() for leaner, quicker inference, plus better NVIDIA Jetson handling with version-aware detection and more robust autobackend behavior. 🚀

📊 Key Changes

  • In-place layer fusion (primary change) ⚡

    • Conv/BatchNorm and Deconv/BatchNorm fusion now happens in-place, updating existing layers directly with gradients disabled.
    • Smarter bias handling (including transposed conv): correctly registers or updates fused bias without creating new layer objects.
    • Result: faster fuse(), lower memory use, and cleaner models. See PR “3x faster inplace model.fuse()” by @​glenn-jocher.

  • Jetson-aware autobackend improvements ✅

    • On Jetson with JetPack 5: move the model to device before fuse() to avoid runtime issues; other devices fuse first, then move to device.
    • attempt_load_weights now performs fuse().eval() before a single, final .to(device) for consistency.
    • is_jetson is imported and cached for faster repeated checks. See PR “Autobackend model.fuse() order of operations speedup” by @​Laughing-q.

  • Version-aware Jetson detection 🔍

  • Version bump

    • version → 8.3.189.

🎯 Purpose & Impact

  • Faster inference setups and smaller memory footprint 🚀

    • In-place fuse() eliminates extra allocations and object creation, accelerating model preparation and inference.

  • More reliable Jetson deployments, especially on JetPack 5 🧩

    • Adjusted fuse/device order fixes fusion-related errors on edge devices, improving stability in production pipelines.

  • Easier, precise hardware targeting 🛠️

    • Version-aware is_jetson() enables conditional logic by JetPack version (e.g., selecting models, dependencies, or optimizations) with a simple call.

  • Backward compatible and safer defaults ✅

    • Existing usage of model.fuse() and is_jetson() continues to work, now faster and more robust.

Quick examples:

  • Faster in-place fusion

    from ultralytics import YOLO

model = YOLO('yolov8n.pt').model
model = model.fuse()  # now runs ~3x faster and in-place

  • Jetson version checks

    from ultralytics.utils import is_jetson

if is_jetson():        # Any Jetson
    ...
if is_jetson(5):       # Specifically JetPack 5
    ...

Happy building and deploying! ✨

What's Changed

Full Changelog: ultralytics/ultralytics@v8.3.188...v8.3.189

v8.3.188: - ultralytics 8.3.188 Faster downloads via a single request (#​21850)

Compare Source

🌟 Summary

Faster, more reliable downloads and smoother developer workflows, plus improved export compatibility and clearer JSON outputs — all wrapped in Ultralytics 8.3.188. 🚀

📊 Key Changes

  • Faster downloads via single request flow and smarter disk-space checks (PR #​21850 by @​glenn-jocher) ⚡
    • Eliminates slow HEAD requests; checks disk only after getting Content-Length.
    • Adaptive buffer size for streaming; precise progress updates.
    • Immediate fail on low disk space to avoid pointless retries.
    • Modernized type hints and cleaner signatures across download utilities.
  • Export compatibility updates (PR #​21758 by @​Y-T-G) 🔄
    • ONNX upper bound removed on Linux/Windows; macOS retains <1.18.0 due to TensorFlow hangs.
    • Updated simplification stack: onnxslim>=0.1.65.
  • JSON outputs include file_name for easier traceability (PR #​21837 by @​olena-hul-dataspan) 🏷️
    • Added "file_name" across RT-DETR, YOLO Detect, and YOLO OBB validators.
  • TQDM robustness improvements (PR #​21849 by @​Laughing-q) 📈
    • Handles total=0 gracefully; richer type hints for better DX.
  • Docs developer UX upgrades (PRs #​21842 by @​glenn-jocher, #​21848 by @​Laughing-q) 🧰
    • Auto-serve docs after build on macOS and Linux with unified logging.
  • Benchmark display polish (PR #​21820 by @​onuralpszr) 🧪
    • Clear, full, and indexed tables with safe null handling.
  • Docs/content touch-ups (PRs #​21835 typo fixes, #​21841 authors metadata) ✨

🎯 Purpose & Impact

  • Speed and reliability: Large model and asset downloads are faster and more robust, improving setup and update times. 🚄
  • Fewer dependency conflicts: Relaxed ONNX constraints reduce install friction on Linux/Windows; macOS remains stable. 🧩
  • Easier result tracking: Including file_name in predictions.json simplifies mapping outputs back to source images. 🔎
  • Better progress feedback: More accurate progress bars and safer behavior for edge cases. ✅
  • Smoother docs workflows: Auto-serve on macOS/Linux and consistent logging streamline local documentation development. 🧑‍💻
  • Cleaner benchmarks: Readable, complete tables help compare formats and performance at a glance. 📊

Upgrade with:
pip install -U ultralytics

What's Changed

New Contributors

Full Changelog: ultralytics/ultralytics@v8.3.187...v8.3.188

v8.3.187: - ultralytics 8.3.187 SAM2: Add SAM2DynamicInteractivePredictor support few-shot inference (#​21232)

Compare Source

🌟 Summary

SAM 2 gets a powerful, training‑free upgrade for interactive, few‑shot multi-object segmentation and tracking, while Ultralytics switches core data operations to Polars for faster, lighter analytics. 🚀

📊 Key Changes

  • SAM2DynamicInteractivePredictor (new, priority)

    • Adds dynamic, interactive few-shot segmentation/tracking across images and video-like sequences.
    • Supports prompts via boxes, points, and masks with real-time memory updates and per-object IDs.
    • Enables continual learning: refine existing objects over time without retraining. 🎯
    • Integrates seamlessly with existing SAM2 models.
    • Docs include examples, API reference, and use cases. See the SAM 2 docs page.

  • Data layer refactor: Pandas ➜ Polars

    • to_df() now returns a Polars DataFrame; CSV/JSON exports use Polars.
    • Removed export helpers: XML, HTML, SQL (and related tests/docs).
    • Training/benchmarks/plotting reading results now use Polars.
    • Dependencies updated: added polars, removed pandas and pandas-stubs.
    • See refactor PR details.

  • Visualization and UX

    • labels.jpg now shows instance counts on class bars for small-class datasets. 📊
    • plot_tune_results() gains consistent Matplotlib styling via @plt_settings() for headless and notebook runs.
    • README/docs banners updated for the YOLOvision event (correct sizing + tracking). 🖼️

  • Documentation improvements

    • SAM 2 docs: detailed guide for dynamic interactive segment-and-track with examples and API reference.
    • OBB task docs: clarify YOLO11 OBB angles must be in [0, 90). 🧭
    • MobileSAM docs: corrected comparison—MobileSAM is “7× smaller, 5× faster” than FastSAM.
    • Docs build reliability improved via mkdocs-ultralytics-plugin>=0.1.29.

🎯 Purpose & Impact

  • Interactive, few-shot segmentation/tracking (SAM2DynamicInteractivePredictor)
    • Purpose: Let users add/track multiple objects over time, refine them interactively, and share memory across frames or independent images—no additional training needed.
    • Impact: Faster video annotation, interactive editing, surveillance, medical/time-series use cases, and semi-automatic dataset labeling with improved consistency. ✨
    • Minimal example: 
      from ultralytics.models.sam import SAM2DynamicInteractivePredictor

overrides = dict(model="sam2_t.pt", task="segment", mode="predict", imgsz=1024, conf=0.01, save=False)
predictor = SAM2DynamicInteractivePredictor(overrides=overrides, max_obj_num=10)

Add an object with a box prompt and store in memory

predictor.inference(img="image1.jpg", bboxes=[[100, 100, 200, 200]], obj_ids=[1], update_memory=True)

Track it in a new image

results = predictor(source="image2.jpg")

Add another object later

predictor.inference(img="image3.jpg", bboxes=[[300, 300, 400, 400]], obj_ids=[2], update_memory=True)

Continue inference

results = predictor(source="image4.jpg")
```

  • Faster, lighter data workflows with Polars

    • Purpose: Improve performance and reduce dependencies across training, validation, benchmarking, and logging.
    • Impact: Quicker data exports and plotting; smaller dependency footprint. Breaking change: code expecting Pandas DataFrames from to_df() must adapt to Polars (e.g., df.to_pandas() if needed). Removed XML/HTML/SQL export helpers—use CSV/JSON or Polars I/O instead.

  • Better plots and docs

    • Purpose: Enhance clarity and reliability of visual outputs and documentation accuracy.
    • Impact: Easier result interpretation (bar labels), more robust plotting in CI/headless environments, clearer OBB labeling constraints, and corrected MobileSAM claims. No changes to training/inference APIs.

Useful links:

  • Read the SAM 2 dynamic interactive docs with examples
  • See the Polars refactor PR
  • Learn about the OBB angle clarification
  • Check the labels.jpg improvement PR
  • Follow the YOLOvision event updates

What's Changed

New Contributors

Full Changelog: ultralytics/ultralytics@v8.3.186...v8.3.187

v8.3.186: - ultralytics 8.3.186 Zero-dependency TQDM progress bars (#​21790)

Compare Source

🌟 Summary

Lean and polished progress bars, sturdier downloads, and smarter GPU handling. v8.3.186 replaces third‑party tqdm with a fast, zero‑dependency TQDM, improves reliability across downloads and exports, and streamlines CI/docs. 🚀

📊 Key Changes

  • Zero‑dependency TQDM progress bars (PR #​21790 by @​glenn-jocher) ✨
    • New ultralytics.utils.tqdm.TQDM with iterator/context support and clean single‑line, rich‑style output.
    • Auto-disables on quiet logs, adapts to terminal width, shows rate/ETA, and throttles in GitHub Actions.
    • tqdm is removed from package dependencies; internal code and docs now use TQDM.
    • New docs reference: utils/tqdm.
    • Minor UX tweak: zip/unzip units now read “files” for clarity.
  • More reliable downloads (PRs #​21791, #​21794) 🔁
    • safe_download() detects partial downloads via Content-Length, logs clear warnings, and cleans incomplete files between retries.
  • Smarter DataLoader memory pinning (PR #​21807) 🧠
    • pin_memory is now enabled only on CUDA systems to avoid unnecessary warnings on CPU-only setups.
  • Export stability fix (PR #​21802) ⚙️
    • Ensures CoreML export tensors are created on the same device as the model (works across CPU/GPU, macOS/Linux).
  • Region counting visualization upgrade (PR #​21805) 🗺️
    • Draws true polygon regions and centers labels at polygon centroids for clearer overlays.
  • GPU monitoring package update (PR #​21795) 🧩
    • Switch from pynvml to NVIDIA’s official nvidia-ml-py package; CI installs it for GPU tests (PR #​21792).
  • Faster installs in notebooks/docs (PR #​21793) ⚡
    • Examples now use !uv pip install ... for speed and reliability.
  • Docs/infra cleanups (PRs #​21797, #​21810 and content/link updates) 🧹
    • Dependency simplification in docs build; downloads badge now links to a richer ClickHouse dashboard.

Minimal example with the new progress bar:

from ultralytics.utils import TQDM

for _ in TQDM(range(100), desc="Processing", unit="items"):
    ...

🎯 Purpose & Impact

  • Lighter installs and fewer external dependencies ➜ smaller footprint, fewer warnings, and more predictable behavior. 📦
  • Cleaner, consistent progress output across terminals and CI, with better rate/ETA reporting. 📈
  • More robust model/dataset downloads; fewer corrupt files and clearer diagnostics on flaky networks. 🌐
  • Reduced “pin_memory” warnings and smarter defaults for CPU/GPU environments. 🧘
  • Fixes export device mismatches for CoreML, improving cross‑platform stability. ✅
  • Better region visualizations improve clarity for zone/region counting solutions. 🎨
  • CI and docs improvements don’t change APIs—no user action required for training, inference, or exports. 🙌

See the Ultralytics Docs for details: https://docs.ultralytics.com

What's Changed

Full Changelog: ultralytics/ultralytics@v8.3.185...v8.3.186

v8.3.185: - Fix TQDM Rich bars to 10 width (#​21789)

Compare Source

🌟 Summary

Improved training progress bars for Rich TQDM users, plus robustness and CI/documentation polish for a smoother overall experience. 🎛️✨

📊 Key Changes

  • Progress bars (priority)
    • Fixed Rich-based TQDM bars to always render clearly: Console width set to 200 and bar width set to 10 when YOLO_TQDM_RICH=true. ✅
    • Resolves missing or truncated bars seen at width 80 in some terminals/CI/Colab.
    • Version bump to 8.3.185.
  • Data loading reliability
    • YOLODataset now defaults to 3 image channels (RGB) if not provided.
    • yolo_bbox2segment() explicitly sets channels=3 to avoid missing-key errors.
  • CI improvements
    • GPU CI job switches to uv for environment management: uv pip install and uv pip list.
    • Removes extra installs (e.g., pynvml, tensorrt, onnxruntime-gpu) to streamline runs.
  • Docs and maintenance
    • Updated link-check workflow to correctly exclude CI files.
    • Clarified SystemLogger.get_metrics() docstring with accurate Python examples.
    • Neural Magic integration links now point to internal docs for a smoother reading experience.

🎯 Purpose & Impact

  • Better UX for progress tracking
    • Rich TQDM users get consistent, readable progress bars across terminals and notebooks. 🙌
    • Default users are unaffected unless YOLO_TQDM_RICH=true is set.
  • More robust workflows
    • Prevents crashes when dataset channel info is missing; safer bbox-to-segmentation conversion. 🛡️
  • Faster, cleaner CI
    • Quicker, more reproducible tests with uv; note some GPU backend tests may have reduced coverage. ⚡
  • Clearer docs
    • Easier navigation and more accurate examples for contributors and users. 📚

Enable Rich progress bars:

  • CLI: YOLO_TQDM_RICH=true yolo train ...
  • Python:
import os
os.environ["YOLO_TQDM_RICH"] = "true"
from ultralytics import YOLO
YOLO("yolo11n.pt").train(data="coco8.yaml", epochs=1)

See the Ultralytics Docs for details: https://docs.ultralytics.com 🚀

What's Changed

Full Changelog: ultralytics/ultralytics@v8.3.184...v8.3.185

v8.3.184: - ultralytics 8.3.184 New SystemLogger class (#​21764)

Compare Source

🌟 Summary

v8.3.184 introduces a new SystemLogger for real-time system metrics and a high-performance ConsoleLogger for early, continuous training log capture — improving observability, reliability, and integrations across YOLO11 workflows. 📈🖥️📝

📊 Key Changes

  • New SystemLogger (ultralytics.utils.logger.SystemLogger)
    • Collects CPU, RAM, disk I/O, network I/O, and NVIDIA GPU stats via pynvml.
    • Integrated into training via platform callbacks; metrics sampled at epoch end.
  • New ConsoleLogger (ultralytics.utils.logger.ConsoleLogger)
    • Captures stdout/stderr and Ultralytics logger output early in training.
    • Deduplicates noisy lines/progress bars and streams to file (train.log) or an API.
    • Enabled via new platform callbacks; logging now starts on trainer initialization.
  • Platform callbacks module
    • ultralytics.utils.callbacks.platform registers start/stop hooks for console capture and metrics collection.
  • Metrics type consistency
    • Checkpoints now save metrics as plain Python floats for better serialization and tool compatibility.
  • Cleaner CI progress bars
    • TQDM updates are throttled on GitHub Actions for cleaner logs; no change for local runs.
  • Documentation improvements
    • API docs now include async functions and a reference for the new NDJSON-to-YOLO converter.
    • New docs pages for platform callbacks and the logger utilities.

🎯 Purpose & Impact

  • Better training visibility and debugging
    • SystemLogger provides actionable telemetry (CPU/GPU/memory/disk/network) to diagnose bottlenecks and environment issues. 🔍
  • More reliable, complete logs
    • ConsoleLogger starts early and deduplicates noise for stable, searchable training logs locally or via APIs. 📜
  • Smoother integrations and exports
    • Float metrics avoid serialization errors when working with Ultralytics HUB, W&B, TensorBoard, JSON/CSV/YAML. 🔗
  • Cleaner CI pipelines
    • Reduced log spam and fewer rate-limit risks on GitHub Actions. ✅
  • Documentation catch-up
    • Clearer API references and discoverability for new utilities and data conversion tools. 📚

Quick start examples:

  • Capture console output to file:
from ultralytics.utils.logger import ConsoleLogger

logger = ConsoleLogger("train.log")
logger.start_capture()
print("Training starts...")
logger.stop_capture()
  • Fetch system metrics on demand:
from ultralytics.utils.logger import SystemLogger

syslog = SystemLogger()
print(syslog.get_metrics())  # {'cpu': ..., 'ram': ..., 'disk': {...}, 'network': {...}, 'gpus': {...}}

See the Ultralytics Docs for details.

What's Changed

Full Changelog: ultralytics/ultralytics@v8.3.183...v8.3.184

v8.3.183: - ultralytics 8.3.183 New NDJSON dataset format support (#​21747)

Compare Source

🌟 Summary

Train YOLO11 models directly from NDJSON datasets with one command — including automatic conversion and image downloading — plus improved YOLOE docs/examples and more reliable GPU CI on GitHub. 🚀

📊 Key Changes

  • NDJSON training support (priority)
    • Built-in NDJSON→YOLO11 converter: convert_ndjson_to_yolo(ndjson_path, output_path=None) by @​glenn-jocher
    • Trainer integration: pass an .ndjson file to model.train(data=...) and it auto-converts and generates data.yaml ✅
    • Fast, resumable downloads: fetches images from URLs with async parallel workers and progress bar ⚡
    • Rich annotation support: handles detection boxes, segments, pose, OBB, and classification; writes the first available type found 🧰
    • Docs updated with NDJSON format examples and advantages
  • YOLOE improvements
    • Corrected pretrained checkpoints and a clearer prompt-free validation example with single_cls=True, plus quieter device selection logs by @​ShuaiLYU 📚
  • CI upgrades
    • Enabled GPU CI on T4 runners, switched to pip for reliability, conditional installs for heavy deps, and adjusted ONNX/TensorRT tests for stability 🔧

Minimal examples:

  • Python: 
    from ultralytics import YOLO
model = YOLO("yolo11n.pt")
model.train(data="path/to/dataset.ndjson", epochs=100, imgsz=640)
  • CLI: 
    yolo detect train data=path/to/dataset.ndjson model=yolo11n.pt epochs=100 imgsz=640

🎯 Purpose & Impact

  • Simpler data onboarding: Use a single NDJSON file (with remote URLs) to kick off training — no manual conversion required. 🧩
  • Faster, scalable pipelines: Async, parallel downloads and streaming-friendly NDJSON make large datasets more manageable. 🚄
  • Broader task coverage: Works across detection, segmentation, pose, OBB, and classification in a unified flow. 🏷️
  • Fewer mistakes: Updated YOLOE examples ensure correct weights and cleaner validation outputs, reducing user friction. ✅
  • More reliable CI: GPU tests run on GitHub with better dependency handling, improving confidence in CUDA/ONNX/TensorRT paths. 🧪

See the Ultralytics Docs for details.

What's Changed

New Contributors

Full Changelog: ultralytics/ultralytics@v8.3.182...v8.3.183

v8.3.182: - ultralytics 8.3.182 Enable type-checking with py.typed marker (#​21698)

Compare Source

🌟 Summary

Ultralytics 8.3.182 declares the package as “partially typed” for better IDE/type-checker support, plus key robustness and usability tweaks for SAM image sizing and mask plotting. 🧩✨

📊 Key Changes

  • Packaging: Added py.typed with a “partial” marker to enable static type checking tools (mypy/pyright) 🏷️
  • SAM reliability: Synced predictor image size with the model and updated internal embedding size calculations to match imgsz changes, preventing shape mismatches 🧠
  • Plotting flexibility: Annotator.masks now accepts NumPy masks and works without a GPU tensor for easier CPU rendering 🎨
  • Docs clarity: imgsz documented as int; clarified rect behavior and how resizing differs for YOLO vs RTDETR models 📚
  • Version bump to 8.3.182 ✅

Tip: You can now plot masks on CPU without providing a torch image tensor:

  • annotator.masks(np_masks, colors, im_gpu=None, alpha=0.5)

See the Ultralytics Docs for details: https://docs.ultralytics.com

🎯 Purpose & Impact

  • Better developer experience: richer IDE autocompletion, fewer type-related false positives, clearer signals that typing is partial 🛠️
  • More robust SAM workflows: avoids feature map mismatches after changing imgsz and stabilizes mask/prompt behavior across batches 🛡️
  • Easier CPU-only visualization: plot segmentation masks directly with NumPy arrays—no GPU tensor required 🚀
  • Clearer training setup: reduced confusion around imgsz and rect across YOLO and RTDETR, leading to more predictable training outcomes 🧭
  • Safe upgrade: no behavioral changes to core training/inference; improvements are non-breaking and quality-of-life focused ✅

What's Changed

Full Changelog: ultralytics/ultralytics@v8.3.181...v8.3.182

v8.3.181: - ultralytics 8.3.181 Support half inference for SAM models (#​21735)

Compare Source

🌟 Summary

v8.3.181 focuses on safer, faster mixed-precision (FP16) support for SAM/SAM2 and YOLOE pipelines, plus reliability fixes across validation exports, source loading, and single-class training. 🚀

📊 Key Changes

  • SAM FP16 support and dtype/device consistency (PR #​21735, priority)

    • Enables half-precision inference for SAM models with consistent dtype handling across blocks, encoders, decoders, utils, and predict paths.
    • Avoids unnecessary float32 casts; ops now respect input tensor dtype (float16/bfloat16 where safe).
    • Predictor now normalizes before casting and sets model dtype based on args.half; unified self.torch_dtype for prompts/masks/buffers.

  • SAM2 robustness without high_res_features (PR #​21726)

    • Decoder gracefully falls back when high-res features are absent; accepts tensor or dict feature inputs.

  • YOLOE device/half propagation and stability (PR #​21670)

    • Predict now forwards device/half flags; prompt tensors follow model precision; softmax casting simplified for consistency.

  • Validation export consistency across tasks (PR #​21719)

    • New scale_preds unifies scaling to original image sizes for detect, OBB, pose, and segment before saving JSON/TXT.

  • YOLOE visual prompt predictor switching fix (PR #​21731)

    • Predictor instance now correctly switches after initialization when using visual prompts.

  • Single-class training compatibility (PR #​21725)

    • Restores classes with single_cls by safely constraining max class index to 0 (no label mutation).

  • CSV source support for inference (PR #​21729)

    • Dataloaders now accept .csv source lists with whitespace-safe parsing.

  • Streamlit Live Inference improvements (PR #​21553)

    • Accepts multiple export formats (.pt, .onnx, .torchscript, .mlpackage, .engine, OpenVINO) and respects full paths provided by users.

  • YOLOE docs enhancements (PR #​21728)

    • Clearer fine-tuning, linear probing, and new export examples; minor classify docs correction.

🎯 Purpose & Impact

  • Faster inference on modern GPUs ⚡
    • FP16 support for SAM reduces memory use and can speed up inference on compatible hardware.
  • Greater numerical stability and fewer dtype/device surprises 🛡️
    • Consistent dtype handling across SAM/SAM2 and YOLOE reduces precision mismatches and unintended casts.
  • More robust segmentation and visual prompting workflows 🧩
    • SAM2 now works even without high-res features; YOLOE honors device/half flags and switches predictors reliably.
  • Accurate and consistent validation exports across tasks 📏
    • JSON/TXT outputs now consistently match original image sizes for detect/OBB/pose/segment.
  • Easier deployment and input management 🧰
    • CSV sources work out of the box; Streamlit Live Inference loads multiple model formats seamlessly.
  • Smoother single-class training ✅
    • Prevents spurious “class exceeds count” errors, improving reliability for single-class projects.

What's Changed

Full Changelog: ultralytics/ultralytics@v8.3.180...v8.3.181

v8.3.180: - ultralytics 8.3.180 new inference_features method for SAM models (#​21708)

Compare Source

🌟 Summary

SAM/SAM2 get a new feature-based inference path with standardized tensor outputs, smarter prompt scaling, and a cleaner API — making segmentation faster to integrate and more consistent end-to-end. 🚀

📊 Key Changes

  • SAM/SAM2: Added inference_features API to run decoding directly from image features (no full image pass required) 🧠⚡
  • Unified outputs: masks and scores now return torch.Tensor (instead of numpy) for consistent PyTorch workflows 🧱
  • Improved prompt handling: _prepare_prompts now takes src_shape and dst_shape for accurate scaling across pipelines 🎯
  • Refactor & cleanup: SAM2 duplicate prompt_inference path removed; shared feature-based logic used for both models 🧹
  • New utilities:
    • _inference_features internal helpers for SAM and SAM2
    • Public inference_features returning masks and xyxy boxes with scores 📦
  • Docs fixes and enhancements:
    • Corrected SAM predictor example labels format 📝
    • Added a Comet ML tutorial video for YOLO11 integration 🎥
  • Downloader improvements: multi-threaded downloads show progress, clearer logs, and no curl dependency in VOC/VisDrone datasets 📥
  • CI/Infra: actions/checkout updated to v5; SBOM generation made more reliable via a dedicated venv 🛠️

🎯 Purpose & Impact

  • Faster pipelines and easier integration: Feature-level inference lets advanced users cache encoder outputs and run multiple prompt decodes without re-encoding images, reducing latency and compute costs 🚄
  • More robust, less brittle code: Standardized tensor outputs simplify downstream processing, batching, and device management ✅
  • Higher accuracy in prompt placement: Explicit source/destination shapes ensure prompts scale correctly, improving segmentation quality across varied image sizes 🎯
  • Cleaner APIs for SAM/SAM2: Reduced duplication and consistent methods make it easier to build on top of SAM models and maintain custom workflows 🧩
  • Better user experience: Clearer docs and examples reduce errors; improved download feedback and reliability save time and headaches ⏱️
  • Stronger CI and compliance posture: Updated actions and SBOM workflow enhance build reliability and security readiness 🔒

What's Changed

Full Changelog: ultralytics/ultralytics@v8.3.179...v8.3.180

Configuration

📅 Schedule: Branch creation - At any time (no schedule defined), Automerge - At any time (no schedule defined).

🚦 Automerge: Enabled.

♻ Rebasing: Whenever PR is behind base branch, or you tick the rebase/retry checkbox.

🔕 Ignore: Close this PR and you won't be reminded about this update again.

  • If you want to rebase/retry this PR, check this box

This PR has been generated by Renovate Bot.

@jazzlyn-bot
jazzlyn-bot Bot force-pushed the renovate/ultralytics-8.x branch from ff3dd10 to c220d4c Compare August 19, 2025 00:43
@jazzlyn-bot
jazzlyn-bot Bot force-pushed the renovate/ultralytics-8.x branch from c220d4c to 8123f9b Compare August 19, 2025 12:16
@jazzlyn-bot jazzlyn-bot Bot changed the title fix(deps): update dependency ultralytics ( 8.3.179 → 8.3.180 ) fix(deps): update dependency ultralytics ( 8.3.179 → 8.3.181 ) Aug 19, 2025
@jazzlyn-bot
jazzlyn-bot Bot force-pushed the renovate/ultralytics-8.x branch from 8123f9b to abe899d Compare August 20, 2025 12:16
@jazzlyn-bot jazzlyn-bot Bot changed the title fix(deps): update dependency ultralytics ( 8.3.179 → 8.3.181 ) fix(deps): update dependency ultralytics ( 8.3.179 → 8.3.182 ) Aug 20, 2025
@jazzlyn-bot
jazzlyn-bot Bot force-pushed the renovate/ultralytics-8.x branch from abe899d to 0aa6a91 Compare August 21, 2025 12:17
@jazzlyn-bot jazzlyn-bot Bot changed the title fix(deps): update dependency ultralytics ( 8.3.179 → 8.3.182 ) fix(deps): update dependency ultralytics ( 8.3.179 → 8.3.183 ) Aug 21, 2025
@jazzlyn-bot
jazzlyn-bot Bot force-pushed the renovate/ultralytics-8.x branch from 0aa6a91 to e1e5b61 Compare August 22, 2025 12:16
@jazzlyn-bot jazzlyn-bot Bot changed the title fix(deps): update dependency ultralytics ( 8.3.179 → 8.3.183 ) fix(deps): update dependency ultralytics ( 8.3.179 → 8.3.184 ) Aug 22, 2025
@jazzlyn-bot
jazzlyn-bot Bot force-pushed the renovate/ultralytics-8.x branch 2 times, most recently from b3b4864 to 05f178e Compare August 25, 2025 00:43
@jazzlyn-bot jazzlyn-bot Bot changed the title fix(deps): update dependency ultralytics ( 8.3.179 → 8.3.184 ) fix(deps): update dependency ultralytics ( 8.3.179 → 8.3.185 ) Aug 25, 2025
@jazzlyn-bot
jazzlyn-bot Bot force-pushed the renovate/ultralytics-8.x branch from 05f178e to 7aef3d8 Compare August 26, 2025 00:41
@jazzlyn-bot jazzlyn-bot Bot changed the title fix(deps): update dependency ultralytics ( 8.3.179 → 8.3.185 ) fix(deps): update dependency ultralytics ( 8.3.179 → 8.3.186 ) Aug 26, 2025
@jazzlyn-bot
jazzlyn-bot Bot force-pushed the renovate/ultralytics-8.x branch from 7aef3d8 to 637ab4b Compare August 27, 2025 12:16
@jazzlyn-bot jazzlyn-bot Bot changed the title fix(deps): update dependency ultralytics ( 8.3.179 → 8.3.186 ) fix(deps): update dependency ultralytics ( 8.3.179 → 8.3.187 ) Aug 27, 2025
@jazzlyn-bot
jazzlyn-bot Bot force-pushed the renovate/ultralytics-8.x branch from 637ab4b to e23ce3f Compare August 28, 2025 12:16
@jazzlyn-bot jazzlyn-bot Bot changed the title fix(deps): update dependency ultralytics ( 8.3.179 → 8.3.187 ) fix(deps): update dependency ultralytics ( 8.3.179 → 8.3.188 ) Aug 28, 2025
@jazzlyn-bot
jazzlyn-bot Bot force-pushed the renovate/ultralytics-8.x branch from e23ce3f to 2f4cd37 Compare August 29, 2025 00:40
@jazzlyn-bot jazzlyn-bot Bot changed the title fix(deps): update dependency ultralytics ( 8.3.179 → 8.3.188 ) fix(deps): update dependency ultralytics ( 8.3.179 → 8.3.189 ) Aug 29, 2025
@jazzlyn-bot
jazzlyn-bot Bot force-pushed the renovate/ultralytics-8.x branch from 2f4cd37 to fdb2863 Compare August 29, 2025 12:16
@jazzlyn-bot
fix(deps): update dependency ultralytics ( 8.3.179 → 8.3.189 )
0fadff1 
| datasource | package     | from    | to      |
| ---------- | ----------- | ------- | ------- |
| pypi       | ultralytics | 8.3.179 | 8.3.189 |
@jazzlyn-bot
jazzlyn-bot Bot force-pushed the renovate/ultralytics-8.x branch from fdb2863 to 0fadff1 Compare August 30, 2025 00:38
@jazzlyn-bot
jazzlyn-bot Bot merged commit 493c730 into main Aug 30, 2025
1 check passed
@jazzlyn-bot
jazzlyn-bot Bot deleted the renovate/ultralytics-8.x branch August 30, 2025 12:13
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

renovate/pep621 renovate/pypi type/patch

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

0 participants