TFLite (original) (raw)

Ultralytics YOLOv5 🚀 is a cutting-edge, state-of-the-art (SOTA) computer vision model developed by Ultralytics. Based on the PyTorch framework, YOLOv5 is renowned for its ease of use, speed, and accuracy. It incorporates insights and best practices from extensive research and development, making it a popular choice for a wide range of vision AI tasks, including object detection, image segmentation, and image classification.

We hope the resources here help you get the most out of YOLOv5. Please browse the YOLOv5 Docs for detailed information, raise an issue on GitHub for support, and join our Discord community for questions and discussions!

To request an Enterprise License, please complete the form at Ultralytics Licensing.

🚀 YOLO11: The Next Evolution

We are excited to announce the launch of Ultralytics YOLO11 🚀, the latest advancement in our state-of-the-art (SOTA) vision models! Available now at the Ultralytics YOLO GitHub repository, YOLO11 builds on our legacy of speed, precision, and ease of use. Whether you're tackling object detection, instance segmentation, pose estimation, image classification, or oriented object detection (OBB), YOLO11 delivers the performance and versatility needed to excel in diverse applications.

Get started today and unlock the full potential of YOLO11! Visit the Ultralytics Docs for comprehensive guides and resources:

Install the ultralytics package

pip install ultralytics

📚 Documentation

See the YOLOv5 Docs for full documentation on training, testing, and deployment. See below for quickstart examples.

Install

Clone the repository and install dependencies in a Python>=3.8.0 environment. Ensure you have PyTorch>=1.8 installed.

Clone the YOLOv5 repository

git clone https://github.com/ultralytics/yolov5

Navigate to the cloned directory

cd yolov5

Install required packages

pip install -r requirements.txt

Inference with PyTorch Hub

Use YOLOv5 via PyTorch Hub for inference. Models are automatically downloaded from the latest YOLOv5 release.

import torch

Load a YOLOv5 model (options: yolov5n, yolov5s, yolov5m, yolov5l, yolov5x)

model = torch.hub.load("ultralytics/yolov5", "yolov5s") # Default: yolov5s

Define the input image source (URL, local file, PIL image, OpenCV frame, numpy array, or list)

img = "https://ultralytics.com/images/zidane.jpg" # Example image

Perform inference (handles batching, resizing, normalization automatically)

results = model(img)

Process the results (options: .print(), .show(), .save(), .crop(), .pandas())

results.print() # Print results to console results.show() # Display results in a window results.save() # Save results to runs/detect/exp

Inference with detect.py

The detect.py script runs inference on various sources. It automatically downloads models from the latest YOLOv5 release and saves the results to the runs/detect directory.

Run inference using a webcam

python detect.py --weights yolov5s.pt --source 0

Run inference on a local image file

python detect.py --weights yolov5s.pt --source img.jpg

Run inference on a local video file

python detect.py --weights yolov5s.pt --source vid.mp4

Run inference on a screen capture

python detect.py --weights yolov5s.pt --source screen

Run inference on a directory of images

python detect.py --weights yolov5s.pt --source path/to/images/

Run inference on a text file listing image paths

python detect.py --weights yolov5s.pt --source list.txt

Run inference on a text file listing stream URLs

python detect.py --weights yolov5s.pt --source list.streams

Run inference using a glob pattern for images

python detect.py --weights yolov5s.pt --source 'path/to/*.jpg'

Run inference on a YouTube video URL

python detect.py --weights yolov5s.pt --source 'https://youtu.be/LNwODJXcvt4'

Run inference on an RTSP, RTMP, or HTTP stream

python detect.py --weights yolov5s.pt --source 'rtsp://example.com/media.mp4'

Training

The commands below demonstrate how to reproduce YOLOv5 COCO dataset results. Both models and datasets are downloaded automatically from the latest YOLOv5 release. Training times for YOLOv5n/s/m/l/x are approximately 1/2/4/6/8 days on a single NVIDIA V100 GPU. Using Multi-GPU training can significantly reduce training time. Use the largest --batch-size your hardware allows, or use --batch-size -1 for YOLOv5 AutoBatch. The batch sizes shown below are for V100-16GB GPUs.

Train YOLOv5n on COCO for 300 epochs

python train.py --data coco.yaml --epochs 300 --weights '' --cfg yolov5n.yaml --batch-size 128

Train YOLOv5s on COCO for 300 epochs

python train.py --data coco.yaml --epochs 300 --weights '' --cfg yolov5s.yaml --batch-size 64

Train YOLOv5m on COCO for 300 epochs

python train.py --data coco.yaml --epochs 300 --weights '' --cfg yolov5m.yaml --batch-size 40

Train YOLOv5l on COCO for 300 epochs

python train.py --data coco.yaml --epochs 300 --weights '' --cfg yolov5l.yaml --batch-size 24

Train YOLOv5x on COCO for 300 epochs

python train.py --data coco.yaml --epochs 300 --weights '' --cfg yolov5x.yaml --batch-size 16

Tutorials

Train Custom Data 🚀 RECOMMENDED: Learn how to train YOLOv5 on your own datasets.
Tips for Best Training Results ☘️: Improve your model's performance with expert tips.
Multi-GPU Training: Speed up training using multiple GPUs.
PyTorch Hub Integration 🌟 NEW: Easily load models using PyTorch Hub.
Model Export (TFLite, ONNX, CoreML, TensorRT) 🚀: Convert your models to various deployment formats like ONNX or TensorRT.
NVIDIA Jetson Deployment 🌟 NEW: Deploy YOLOv5 on NVIDIA Jetson devices.
Test-Time Augmentation (TTA): Enhance prediction accuracy with TTA.
Model Ensembling: Combine multiple models for better performance.
Model Pruning/Sparsity: Optimize models for size and speed.
Hyperparameter Evolution: Automatically find the best training hyperparameters.
Transfer Learning with Frozen Layers: Adapt pretrained models to new tasks efficiently using transfer learning.
Architecture Summary 🌟 NEW: Understand the YOLOv5 model architecture.
Ultralytics HUB Training 🚀 RECOMMENDED: Train and deploy YOLO models using Ultralytics HUB.
ClearML Logging: Integrate with ClearML for experiment tracking.
Neural Magic DeepSparse Integration: Accelerate inference with DeepSparse.
Comet Logging 🌟 NEW: Log experiments using Comet ML.

🧩 Integrations

Our key integrations with leading AI platforms extend the functionality of Ultralytics' offerings, enhancing tasks like dataset labeling, training, visualization, and model management. Discover how Ultralytics, in collaboration with partners like Weights & Biases, Comet ML, Roboflow, and Intel OpenVINO, can optimize your AI workflow. Explore more at Ultralytics Integrations.

Ultralytics HUB 🌟	Weights & Biases	Comet	Neural Magic
Streamline YOLO workflows: Label, train, and deploy effortlessly with Ultralytics HUB. Try now!	Track experiments, hyperparameters, and results with Weights & Biases.	Free forever, Comet ML lets you save YOLO models, resume training, and interactively visualize predictions.	Run YOLO inference up to 6x faster with Neural Magic DeepSparse.

⭐ Ultralytics HUB

Experience seamless AI development with Ultralytics HUB ⭐, the ultimate platform for building, training, and deploying computer vision models. Visualize datasets, train YOLOv5 and YOLOv8 🚀 models, and deploy them to real-world applications without writing any code. Transform images into actionable insights using our cutting-edge tools and user-friendly Ultralytics App. Start your journey for Free today!

🤔 Why YOLOv5?

YOLOv5 is designed for simplicity and ease of use. We prioritize real-world performance and accessibility.

YOLOv5-P5 640 Figure

Figure Notes

COCO AP val denotes the mean Average Precision (mAP) at Intersection over Union (IoU) thresholds from 0.5 to 0.95, measured on the 5,000-image COCO val2017 dataset across various inference sizes (256 to 1536 pixels).
GPU Speed measures the average inference time per image on the COCO val2017 dataset using an AWS p3.2xlarge V100 instance with a batch size of 32.
EfficientDet data is sourced from the google/automl repository at batch size 8.
Reproduce these results using the command: python val.py --task study --data coco.yaml --iou 0.7 --weights yolov5n6.pt yolov5s6.pt yolov5m6.pt yolov5l6.pt yolov5x6.pt

Pretrained Checkpoints

This table shows the performance metrics for various YOLOv5 models trained on the COCO dataset.

Model	Size(pixels)	mAPval50-95	mAPval50	SpeedCPU b1(ms)	SpeedV100 b1(ms)	SpeedV100 b32(ms)	Params(M)	FLOPs@640 (B)
YOLOv5n	640	28.0	45.7	45	6.3	0.6	1.9	4.5
YOLOv5s	640	37.4	56.8	98	6.4	0.9	7.2	16.5
YOLOv5m	640	45.4	64.1	224	8.2	1.7	21.2	49.0
YOLOv5l	640	49.0	67.3	430	10.1	2.7	46.5	109.1
YOLOv5x	640	50.7	68.9	766	12.1	4.8	86.7	205.7
YOLOv5n6	1280	36.0	54.4	153	8.1	2.1	3.2	4.6
YOLOv5s6	1280	44.8	63.7	385	8.2	3.6	12.6	16.8
YOLOv5m6	1280	51.3	69.3	887	11.1	6.8	35.7	50.0
YOLOv5l6	1280	53.7	71.3	1784	15.8	10.5	76.8	111.4
YOLOv5x6+ [TTA]	12801536	55.055.8	72.772.7	3136-	26.2-	19.4-	140.7-	209.8-

Table Notes

All checkpoints were trained for 300 epochs using default settings. Nano (n) and Small (s) models use hyp.scratch-low.yaml hyperparameters, while Medium (m), Large (l), and Extra-Large (x) models use hyp.scratch-high.yaml.
mAPval values represent single-model, single-scale performance on the COCO val2017 dataset.
Reproduce using: python val.py --data coco.yaml --img 640 --conf 0.001 --iou 0.65
Speed metrics are averaged over COCO val images using an AWS p3.2xlarge V100 instance. Non-Maximum Suppression (NMS) time (~1 ms/image) is not included.
Reproduce using: python val.py --data coco.yaml --img 640 --task speed --batch 1
TTA (Test Time Augmentation) includes reflection and scale augmentations for improved accuracy.
Reproduce using: python val.py --data coco.yaml --img 1536 --iou 0.7 --augment

🖼️ Segmentation

The YOLOv5 release v7.0 introduced instance segmentation models that achieve state-of-the-art performance. These models are designed for easy training, validation, and deployment. For full details, see the Release Notes and explore the YOLOv5 Segmentation Colab Notebook for quickstart examples.

Segmentation Checkpoints

YOLOv5 segmentation models were trained on the COCO dataset for 300 epochs at an image size of 640 pixels using A100 GPUs. Models were exported to ONNX FP32 for CPU speed tests and TensorRT FP16 for GPU speed tests. All speed tests were conducted on Google Colab Pro notebooks for reproducibility.

Model	Size(pixels)	mAPbox50-95	mAPmask50-95	Train Time300 epochsA100 (hours)	SpeedONNX CPU(ms)	SpeedTRT A100(ms)	Params(M)	FLOPs@640 (B)
YOLOv5n-seg	640	27.6	23.4	80:17	62.7	1.2	2.0	7.1
YOLOv5s-seg	640	37.6	31.7	88:16	173.3	1.4	7.6	26.4
YOLOv5m-seg	640	45.0	37.1	108:36	427.0	2.2	22.0	70.8
YOLOv5l-seg	640	49.0	39.9	66:43 (2x)	857.4	2.9	47.9	147.7
YOLOv5x-seg	640	50.7	41.4	62:56 (3x)	1579.2	4.5	88.8	265.7

All checkpoints were trained for 300 epochs using the SGD optimizer with lr0=0.01 and weight_decay=5e-5 at an image size of 640 pixels, using default settings.
Training runs are logged at https://wandb.ai/glenn-jocher/YOLOv5_v70_official.
Accuracy values represent single-model, single-scale performance on the COCO dataset.
Reproduce using: python segment/val.py --data coco.yaml --weights yolov5s-seg.pt
Speed metrics are averaged over 100 inference images using a Colab Pro A100 High-RAM instance. Values indicate inference speed only (NMS adds approximately 1ms per image).
Reproduce using: python segment/val.py --data coco.yaml --weights yolov5s-seg.pt --batch 1
Export to ONNX (FP32) and TensorRT (FP16) was performed using export.py.
Reproduce using: python export.py --weights yolov5s-seg.pt --include engine --device 0 --half Segmentation Usage Examples

Train

YOLOv5 segmentation training supports automatic download of the COCO128-seg dataset via the --data coco128-seg.yaml argument. For the full COCO-segments dataset, download it manually using bash data/scripts/get_coco.sh --train --val --segments and then train with python train.py --data coco.yaml.

Train on a single GPU

python segment/train.py --data coco128-seg.yaml --weights yolov5s-seg.pt --img 640

Train using Multi-GPU Distributed Data Parallel (DDP)

python -m torch.distributed.run --nproc_per_node 4 --master_port 1 segment/train.py --data coco128-seg.yaml --weights yolov5s-seg.pt --img 640 --device 0,1,2,3

Val

Validate the mask mean Average Precision (mAP) of YOLOv5s-seg on the COCO dataset:

Download COCO validation segments split (780MB, 5000 images)

bash data/scripts/get_coco.sh --val --segments

Validate the model

python segment/val.py --weights yolov5s-seg.pt --data coco.yaml --img 640

Predict

Use the pretrained YOLOv5m-seg.pt model to perform segmentation on bus.jpg:

Run prediction

python segment/predict.py --weights yolov5m-seg.pt --source data/images/bus.jpg

Load model from PyTorch Hub (Note: Inference support might vary)

model = torch.hub.load("ultralytics/yolov5", "custom", "yolov5m-seg.pt")

Export

Export the YOLOv5s-seg model to ONNX and TensorRT formats:

Export model

python export.py --weights yolov5s-seg.pt --include onnx engine --img 640 --device 0

🏷️ Classification

YOLOv5 release v6.2 introduced support for image classification model training, validation, and deployment. Check the Release Notes for details and the YOLOv5 Classification Colab Notebook for quickstart guides.

Classification Checkpoints

YOLOv5-cls classification models were trained on ImageNet for 90 epochs using a 4xA100 instance. ResNet and EfficientNet models were trained alongside under identical settings for comparison. Models were exported to ONNX FP32 (CPU speed tests) and TensorRT FP16 (GPU speed tests). All speed tests were run on Google Colab Pro for reproducibility.

Model	Size(pixels)	Acctop1	Acctop5	Training90 epochs4xA100 (hours)	SpeedONNX CPU(ms)	SpeedTensorRT V100(ms)	Params(M)	FLOPs@224 (B)
YOLOv5n-cls	224	64.6	85.4	7:59	3.3	0.5	2.5	0.5
YOLOv5s-cls	224	71.5	90.2	8:09	6.6	0.6	5.4	1.4
YOLOv5m-cls	224	75.9	92.9	10:06	15.5	0.9	12.9	3.9
YOLOv5l-cls	224	78.0	94.0	11:56	26.9	1.4	26.5	8.5
YOLOv5x-cls	224	79.0	94.4	15:04	54.3	1.8	48.1	15.9
ResNet18	224	70.3	89.5	6:47	11.2	0.5	11.7	3.7
ResNet34	224	73.9	91.8	8:33	20.6	0.9	21.8	7.4
ResNet50	224	76.8	93.4	11:10	23.4	1.0	25.6	8.5
ResNet101	224	78.5	94.3	17:10	42.1	1.9	44.5	15.9
EfficientNet_b0	224	75.1	92.4	13:03	12.5	1.3	5.3	1.0
EfficientNet_b1	224	76.4	93.2	17:04	14.9	1.6	7.8	1.5
EfficientNet_b2	224	76.6	93.4	17:10	15.9	1.6	9.1	1.7
EfficientNet_b3	224	77.7	94.0	19:19	18.9	1.9	12.2	2.4

Table Notes (click to expand)

All checkpoints were trained for 90 epochs using the SGD optimizer with lr0=0.001 and weight_decay=5e-5 at an image size of 224 pixels, using default settings.
Training runs are logged at https://wandb.ai/glenn-jocher/YOLOv5-Classifier-v6-2.
Accuracy values (top-1 and top-5) represent single-model, single-scale performance on the ImageNet-1k dataset.
Reproduce using: python classify/val.py --data ../datasets/imagenet --img 224
Speed metrics are averaged over 100 inference images using a Google Colab Pro V100 High-RAM instance.
Reproduce using: python classify/val.py --data ../datasets/imagenet --img 224 --batch 1
Export to ONNX (FP32) and TensorRT (FP16) was performed using export.py.
Reproduce using: python export.py --weights yolov5s-cls.pt --include engine onnx --imgsz 224 Classification Usage Examples

Train

YOLOv5 classification training supports automatic download for datasets like MNIST, Fashion-MNIST, CIFAR10, CIFAR100, Imagenette, Imagewoof, and ImageNet using the --data argument. For example, start training on MNIST with --data mnist.

Train on a single GPU using CIFAR-100 dataset

python classify/train.py --model yolov5s-cls.pt --data cifar100 --epochs 5 --img 224 --batch 128

Train using Multi-GPU DDP on ImageNet dataset

python -m torch.distributed.run --nproc_per_node 4 --master_port 1 classify/train.py --model yolov5s-cls.pt --data imagenet --epochs 5 --img 224 --device 0,1,2,3

Val

Validate the accuracy of the YOLOv5m-cls model on the ImageNet-1k validation dataset:

Download ImageNet validation split (6.3GB, 50,000 images)

bash data/scripts/get_imagenet.sh --val

Validate the model

python classify/val.py --weights yolov5m-cls.pt --data ../datasets/imagenet --img 224

Predict

Use the pretrained YOLOv5s-cls.pt model to classify the image bus.jpg:

Run prediction

python classify/predict.py --weights yolov5s-cls.pt --source data/images/bus.jpg

Load model from PyTorch Hub

model = torch.hub.load("ultralytics/yolov5", "custom", "yolov5s-cls.pt")

Export

Export trained YOLOv5s-cls, ResNet50, and EfficientNet_b0 models to ONNX and TensorRT formats:

Export models

python export.py --weights yolov5s-cls.pt resnet50.pt efficientnet_b0.pt --include onnx engine --img 224

☁️ Environments

Get started quickly with our pre-configured environments. Click the icons below for setup details.

🤝 Contribute

We welcome your contributions! Making YOLOv5 accessible and effective is a community effort. Please see our Contributing Guide to get started. Share your feedback through the YOLOv5 Survey. Thank you to all our contributors for making YOLOv5 better!

📜 License

Ultralytics provides two licensing options to meet different needs:

AGPL-3.0 License: An OSI-approved open-source license ideal for academic research, personal projects, and testing. It promotes open collaboration and knowledge sharing. See the LICENSE file for details.
Enterprise License: Tailored for commercial applications, this license allows seamless integration of Ultralytics software and AI models into commercial products and services, bypassing the open-source requirements of AGPL-3.0. For commercial use cases, please contact us via Ultralytics Licensing.

📧 Contact

For bug reports and feature requests related to YOLOv5, please visit GitHub Issues. For general questions, discussions, and community support, join our Discord server!