[PDF] YOLOv10: Real-Time End-to-End Object Detection | Semantic Scholar (2024)

[PDF] Semantic Reader

This work introduces a stepwise spatial global-local aggregation network that combines the local information from the neighboring frames and global semantics from the current frame to eliminate the feature degradation in video object detection.

This paper proposes the Real-Time DEtection TRansformer (RT-DETR), the first real-time end-to-end object detector to the best knowledge that addresses the above dilemma and designs an efficient hybrid encoder to expeditiously process multi-scale features by decoupling intra-scale interaction and cross-scale fusion to improve speed.

DAMO-YOLO is extended from YOLO with some new technologies, including Neural Architecture Search (NAS), efficient Reparameterized Generalized-FPN (RepGFPN), a lightweight head with AlignedOTA label assignment, and distillation enhancement to improve performance to a higher level.

This study provides an advanced Gatherand-Distribute mechanism (GD) mechanism, which is realized with convolution and self-attention operations, and implements MAE-style pretraining in the YOLO-series for the first time, allowing Y OLOseries models could be to benefit from unsupervised pretraining.

An efficient real-time object detector is designed that exceeds the YOLO series and is easily extensible for many object recognition tasks such as instance segmentation and rotated object detection and introduces soft labels when calculating matching costs in the dynamic label assignment to improve accuracy.

The core design is based on a series of investigations on how convolutions with different kernel sizes affect the detection performance of objects at different scales, resulting in a new strategy that can strongly enhance multi-scale feature representations of real-time object detectors.

A Prediction-aware One-To-One (POTO) label assignment for classification is introduced to enable end-to-end detection, which obtains comparable performance with NMS and a simple 3D Max Filtering is proposed to utilize the multi-scale features and improve the discriminability of convolutions in the local region.

It is pointed out that one-to-one positive sample assignment is the key factor, while, one- to-many assignment in previous detectors causes redundant predictions in inference, and the concept of score gap is introduced to explore the effect of matching cost.

This work presents a new method that views object detection as a direct set prediction problem, and demonstrates accuracy and run-time performance on par with the well-established and highly-optimized Faster RCNN baseline on the challenging COCO object detection dataset.

MOTR is proposed, which extends DETR and introduces track query to model the tracked instances in the entire video to enhance temporal relation modeling and serve as a stronger baseline for future research on temporal modeling and Transformer-based trackers.

Compared to state-of-the-art detection systems, YOLO makes more localization errors but is less likely to predict false positives on background, and outperforms other detection methods, including DPM and R-CNN, when generalizing from natural images to other domains like artwork.

...

...