⭐ Support LibreYOLO. The best way to help is to star the repo. Feel free to open an issue if you encounter problems or have suggestions, and code contributions are very welcome (see CONTRIBUTING.md). If you want to contribute economically, that is very welcome: see the sponsorship program.
An MIT-licensed computer vision library. Detection, segmentation, pose, depth, OCR and a dozen more tasks behind one small API, with training and export included rather than sold separately. Reads common YOLO-format datasets, so existing workflows port over with minimal changes.
pip install libreyolofrom libreyolo import LibreYOLO, SAMPLE_IMAGE
model = LibreYOLO("LibreYOLO9t.pt")
result = model(SAMPLE_IMAGE, save=True)Optional extras
The base install covers YOLOv9 and the other core detectors, training, and
inference. Add an extra when you need a heavier family or an export backend.
Comma-separate to combine, for example pip install "libreyolo[rfdetr,hub-kernels]".
| Group | Extras |
|---|---|
| Export | onnx, tensorrt, openvino, coreml, coreai, tflite (alias litert), ncnn, mnn, paddle, executorch |
| Serving | triton |
| Models | rfdetr, vlm, sam, openvocab, clip, siglip2, eomt, midas, modus, sensenova, gaze |
| Training | lora, plots, tensorboard, mlflow, wandb, comet, clearml, neptune, dvclive |
| Speed | fast-eval, hub-kernels |
| Sources | stream |
| Everything | pip install "libreyolo[all]" |
executorch, coreai and neptune are deliberately left out of all: they
pin torch or protobuf in ways that would drag the rest of the environment with
them. Full list and per-backend notes in the
install guide.
Install from source
git clone https://github.com/LibreYOLO/libreyolo.git
cd libreyolo
pip install -e .A plain clone checks out release, the stable branch matching the published
package. For unreleased work, git checkout dev.
The same three lines run every task. Only the checkpoint changes.
from libreyolo import LibreYOLO
LibreYOLO("LibreYOLO9t.pt")("street.jpg", save=True) # detection
LibreYOLO("LibreDeepLabv3mv3-sem.pt")("street.jpg", save=True) # semantic segmentation
LibreYOLO("LibreHRNetw32-pose.pt")("street.jpg", save=True) # pose
LibreYOLO("LibreMiDaSs-depth.pt")("street.jpg", save=True) # depth
LibreYOLO("LibreFeyNobgl-matte.pt")("portrait.jpg", save=True) # background removal
LibreYOLO("LibreRTDETRv2n-obb.pt")("aerial.jpg", save=True) # oriented boxesSources are not just files. Point it at a webcam, an RTSP stream, a video, a directory, a YouTube URL or your screen:
libreyolo predict --model yolo9-t --source 0 --show # webcam
libreyolo predict --model yolo9-t --source rtsp://camera/1 # network camera
libreyolo predict --model yolo9-t --source screen # screen capture| Task | Models |
|---|---|
| Detection | YOLOv9, RF-DETR, YOLOX, YOLO-NAS, D-FINE, DEIM, RT-DETR v1/v2/v4, RTMDet, PicoDet, PP-YOLOE, YOLOv7, EfficientDet, and the classics: DETR, Deformable DETR, DINO-DETR, LW-DETR, Faster R-CNN, RetinaNet, SSD, FCOS, CenterNet |
| Tiny objects | TinyFormer, Dome-DETR (aerial, drone, remote sensing) |
| Instance segmentation | RF-DETR, RTMDet, D-FINE, Mask R-CNN |
| Promptable segmentation | SAM, SAM 2, SAM 3, MobileSAM, EdgeTAM, PicoSAM3 |
| Semantic segmentation | SegFormer, PIDNet, PP-LiteSeg, U-Net, DeepLabv3, FCN, LingBot-Vision, DINOv2, EoMT |
| Panoptic segmentation | EoMT |
| Pose | RF-DETR, YOLO-NAS, HRNet, DEKR, EC |
| Oriented boxes | RF-DETR, RT-DETRv2, YOLO-NAS-R |
| 3D detection | WildDet3D, 3D-MOOD, FCOS3D, DetAny3D (separate runtime) |
| Classification | MobileNetV4, ConvNeXt, EfficientNetV2, ResNet, ViT, Swin, DeiT, VGG, AlexNet, CLIP, SigLIP2, DINOv2 |
| Depth | Depth Anything 3, Depth Anything V2, ZipDepth, MiDaS |
| Surface normals | MoGe-2 |
| Depth, normals and albedo | Marigold V2 |
| Edges | DexiNed, TEED |
| Embeddings | LibreFaceEmbedder, CLIP, SigLIP2, Perception Encoder (image, text, whole-video; also zero-shot classify), DINOv2 |
| Video embeddings | V-JEPA 2 (clip embedding and trainable classification probe), LeVJEPA (clip and patch embeddings; CC BY-NC 4.0 weights) |
| Body mesh | SAM 3D Body |
| Restoration | DDColor, HVI-CIDNet, LaMa, NAFNet, QuickSRNet, Real-ESRGAN, SwinIR |
| Background removal | BiRefNet, FeyNobg, BEN2, ViTMatte |
| OCR | PP-OCR |
| Point detection | FOMO, LocateAnything |
| Gaze | L2CS |
| Open vocabulary and VLMs | Grounding DINO, OWLv2, OmDet-Turbo, OV-DEIM, Florence-2, Kosmos-2, Qwen3-VL, InternVL3, LFM2-VL, North Micro Vision, SmolVLM2, Gemma 4, Moondream, Molmo2, MODUS |
| Robot actions (VLA) | SmolVLA, ACT, Diffusion Policy |
Per-family sizes, checkpoints and parity evidence live in the model reference.
from libreyolo import LibreYOLO
model = LibreYOLO("LibreYOLO9t.pt")
model.train(data="dataset.yaml", epochs=100, imgsz=640)libreyolo train --model yolo9-t --data dataset.yaml --epochs 100Multi-GPU, LoRA, layer freezing, distillation, from-scratch training, and TensorBoard, MLflow, Weights & Biases, Comet, ClearML, Neptune and DVCLive logging are all supported. See the training guide.
Twelve formats: ONNX, TorchScript, TensorRT, OpenVINO, CoreML, Core AI, TFLite (LiteRT), NCNN, MNN, RKNN, Paddle and ExecuTorch. Plus NVIDIA Triton serving and DeepStream config generation.
libreyolo export --model yolo9-t --format onnxSupport varies by family and task, see the export matrix.
- Docs covers install, tasks, models, training, prediction, export and the CLI
- Benchmarks for independent numbers
- CHANGELOG.md for what changed
The code is written in free time. Training new models, such as LibreYOLO26 and MIT retrains of families with restrictive weights, costs rented GPU time, and the LibreYOLO Sponsorship Program pays for it. Companies get their logo here from $100 a month, and from $500 a month are named in every release and in the model card of every weight released while they sponsor. Hardware manufacturers can sponsor by sending a device.
No sponsors yet.
- Code: MIT License.
- Weights: pre-trained weights may inherit licensing from their original source, and not all of them are permissive. Check the license on the specific Hugging Face repo before you use one commercially. Every LibreYOLO Hugging Face model states its license.
