YOLOv10: Real-Time End-to-End Object Detection#

YOLOv10, released in May 2024 and built on the Ultralytics Python package by researchers at Tsinghua University, introduces a new approach to real-time object detection, addressing both the post-processing and model architecture deficiencies found in previous YOLO versions. By eliminating non-maximum suppression (NMS) and optimizing various model components, YOLOv10 achieved strong performance with significantly reduced computational overhead at its time of release. Its NMS-free end-to-end design pioneered an approach that has been further developed in YOLO26.

Watch: How to Train YOLOv10 on SKU-110k Dataset using Ultralytics | Retail Dataset

Overview#

Real-time object detection aims to accurately predict object categories and positions in images with low latency. The YOLO series has been at the forefront of this research due to its balance between performance and efficiency. However, reliance on NMS and architectural inefficiencies have hindered optimal performance. YOLOv10 addresses these issues by introducing consistent dual assignments for NMS-free training and a holistic efficiency-accuracy driven model design strategy.

Architecture#

The architecture of YOLOv10 builds upon the strengths of previous YOLO models while introducing several key innovations. The model architecture consists of the following components:

1. Backbone: Responsible for feature extraction, the backbone in YOLOv10 uses an enhanced version of CSPNet (Cross Stage Partial Network) to improve gradient flow and reduce computational redundancy.
2. Neck: The neck is designed to aggregate features from different scales and passes them to the head. It includes PAN (Path Aggregation Network) layers for effective multiscale feature fusion.
3. One-to-Many Head: Generates multiple predictions per object during training to provide rich supervisory signals and improve learning accuracy.
4. One-to-One Head: Generates a single best prediction per object during inference to eliminate the need for NMS, thereby reducing latency and improving efficiency.

Key Features#

1. NMS-Free Training: Utilizes consistent dual assignments to eliminate the need for NMS, reducing inference latency.
2. Holistic Model Design: Comprehensive optimization of various components from both efficiency and accuracy perspectives, including lightweight classification heads, spatial-channel decoupled downsampling, and rank-guided block design.
3. Enhanced Model Capabilities: Incorporates large-kernel convolutions and partial self-attention modules to improve performance without significant computational cost.

Model Variants#

YOLOv10 comes in various model scales to cater to different application needs:

• YOLOv10n: Nano version for extremely resource-constrained environments.
• YOLOv10s: Small version balancing speed and accuracy.
• YOLOv10m: Medium version for general-purpose use.
• YOLOv10b: Balanced version with increased width for higher accuracy.
• YOLOv10l: Large version for higher accuracy at the cost of increased computational resources.
• YOLOv10x: Extra-large version for maximum accuracy and performance.

Performance#

At release, YOLOv10 outperformed previous YOLO versions and other state-of-the-art models in accuracy and efficiency. For example, YOLOv10s is 1.8x faster than RT-DETR-R18 with similar AP on the COCO dataset, and YOLOv10b has 46% less latency and 25% fewer parameters than YOLOv9-C with the same performance.

Methodology#

Consistent Dual Assignments for NMS-Free Training#

YOLOv10 employs dual label assignments, combining one-to-many and one-to-one strategies during training to ensure rich supervision and efficient end-to-end deployment. Ultralytics prediction and validation default to the one-to-many head with NMS; set nms=False to select the NMS-free head. The consistent matching metric aligns the supervision between both strategies, enhancing the quality of predictions during inference.

Holistic Efficiency-Accuracy Driven Model Design#

Efficiency Enhancements#

1. Lightweight Classification Head: Reduces the computational overhead of the classification head by using depth-wise separable convolutions.
2. Spatial-Channel Decoupled Downsampling: Decouples spatial reduction and channel modulation to minimize information loss and computational cost.
3. Rank-Guided Block Design: Adapts block design based on intrinsic stage redundancy, ensuring optimal parameter utilization.

Accuracy Enhancements#

1. Large-Kernel Convolution: Enlarges the receptive field to enhance feature extraction capability.
2. Partial Self-Attention (PSA): Incorporates self-attention modules to improve global representation learning with minimal overhead.

Experiments and Results#

YOLOv10 was evaluated on standard benchmarks such as COCO. At release, every variant improved the latency-accuracy trade-off compared to previous YOLO versions and other contemporary detectors.

Comparisons#

Compared to other state-of-the-art detectors:

• YOLOv10s / x are 1.8× / 1.3× faster than RT-DETR-R18 / R101 with similar accuracy
• YOLOv10b has 25% fewer parameters and 46% lower latency than YOLOv9-C at same accuracy
• YOLOv10l / x outperform YOLOv8l / x by 0.3 AP / 0.5 AP with 1.8× / 2.3× fewer parameters

Params and FLOPs values are for the fused model after model.fuse(), which merges Conv and BatchNorm layers and removes the auxiliary one-to-many detection head. Pretrained checkpoints retain the full training architecture and may show higher counts.

Usage Examples#

Models can also be trained on cloud GPUs through Ultralytics Platform. To predict on new images with YOLOv10:

from ultralytics import YOLO

# Load a pretrained YOLOv10n model
model = YOLO("yolov10n.pt")

# Perform object detection on an image
results = model("image.jpg")

# Display the results
results[0].show()

For training YOLOv10 on a custom dataset:

from ultralytics import YOLO

# Load YOLOv10n model from scratch
model = YOLO("yolov10n.yaml")

# Train the model
model.train(data="coco8.yaml", epochs=100, imgsz=640)

Supported Tasks and Modes#

The YOLOv10 model series offers a range of models, each optimized for object detection. The variants cover a range of computational budgets and accuracy requirements.

Exporting YOLOv10#

Due to the new operations introduced with YOLOv10, not all export formats provided by Ultralytics are currently supported. The following table lists the formats that Ultralytics has successfully converted for YOLOv10. Pull requests that add export support for additional formats are welcome.

Conclusion#

At its release, YOLOv10 advanced real-time object detection by removing NMS post-processing and redesigning model components for both efficiency and accuracy. Its NMS-free approach pioneered end-to-end object detection in the YOLO family. YOLO26, the latest Ultralytics model, builds on this with improved performance and NMS-free inference.

Citations and Acknowledgments#

We would like to acknowledge the YOLOv10 authors from Tsinghua University for their extensive research and significant contributions to the Ultralytics framework:

@inproceedings{wang2024yolov10,
  title={YOLOv10: Real-Time End-to-End Object Detection},
  author={Wang, Ao and Chen, Hui and Liu, Lihao and Chen, Kai and Lin, Zijia and Han, Jungong and Ding, Guiguang},
  booktitle={Advances in Neural Information Processing Systems},
  doi = {10.52202/079017-3429},
  url = {https://proceedings.neurips.cc/paper_files/paper/2024/file/c34ddd05eb089991f06f3c5dc36836e0-Paper-Conference.pdf},
  volume={37},
  pages={107984--108011},
  year={2024}
}

For detailed implementation, architectural innovations, and experimental results, please refer to the YOLOv10 research paper and GitHub repository by the Tsinghua University team.

FAQ#