YOLOX: Exceeding YOLO Series in 2021
By Zheng Ge, Songtao Liu, Feng Wang, Zeming Li, Jian‐Yuan Sun. arXiv.org, 2021
6,005 citations recorded by the source corpus.
In this report, we present some experienced improvements to YOLO series, forming a new high-performance detector -- YOLOX. We switch the YOLO detector to an anchor-free manner and conduct other advanced detection techniques, i.e., a decoupled head and the leading label assignment strategy SimOTA to achieve state-of-the-art results across a large scale range of models: For YOLO-Nano with only 0.91M parameters and 1.08G FLOPs, we get 25.3% AP on COCO, surpassing NanoDet by 1.8% AP; for YOLOv3, one of the most widely used detectors in industry, we boost it to 47.3% AP on COCO, outperforming the current best practice by 3.0% AP; for YOLOX-L with roughly the same amount of parameters as YOLOv4-CSP, YOLOv5-L, we achieve 50.0% AP on COCO at a speed of 68.9 FPS on Tesla V100, exceeding YOLOv5-L by 1.8% AP. Further, we won the 1st Place on Streaming Perception Challenge (Workshop on Autonomous Driving at CVPR 2021) using a single YOLOX-L model. We hope this report can provide useful experience for developers and researchers in practical scenes, and we also provide deploy versions with ONNX, TensorRT, NCNN, and Openvino supported. Source code is at https://github.com/Megvii-BaseDetection/YOLOX.
Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Radiology, Nuclear Medicine and Imaging
View Paper: https://www.semanticscholar.org/paper/c01b385205e488a731c8c8c11c0c494d426beb03
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
In Computer Vision and Pattern Recognition
- Dacheng Tao
- Luc Van Gool
- Ming-Hsuan Yang
- Xiaogang Wang
- Lei Zhang
- Trevor Darrell
- Jitendra Malik
- Chunhua Shen
In Radiology, Nuclear Medicine and Imaging
Citations in the corpus
- Observation-Centric SORT: Rethinking SORT for Robust Multi-Object Tracking (2022, methodology): “We train to use the YOLOX [20] trained from the MOT17 training set to provide detections on DanceTrack and find the tracking performance of OC-SORT is already higher than baselines.”
- YOLO-Pose: Enhancing YOLO for Multi Person Pose Estimation Using Object Keypoint Similarity Loss (2022, methodology): “Since our enhancement is independent of anchor width and height, it can be easily extended to anchor-free object detection approaches like YOLOX [2], FCOS [28].”
- YOLOv7: Trainable Bag-of-Freebies Sets New State-of-the-Art for Real-Time Object Detectors (2022, methodology): “YOLOv7-E6 object detector (56 FPS V100, 55.9% AP) outperforms both transformer-based detector SWIN-L Cascade-Mask R-CNN (9.2 FPS A100, 53.9% AP) by 509% in speed and 2% in accuracy, and convolutional-based detector ConvNeXt-XL Cascade-Mask R-CNN (8.6 FPS A100, 55.2% AP) by 551% in speed and 0.7% AP ”
- Fast Segment Anything (2023, methodology): “The architecture of YOLOv8 [16] develops from its predecessor, YOLOv5 [15], integrating key design aspects from recent algorithms such as YOLOX [10], YOLOv6 [22], and YOLOv7 [35].”
- CLRNet: Cross Layer Refinement Network for Lane Detection (2022, background): “During training, each ground truth lane is assigned with one or more predicted lanes dynamically as positive samples, which is inspired by [4].”
- DiffusionDet: Diffusion Model for Object Detection (2022, methodology): “We assign multiple predictions to each ground truth by selecting the top k predictions with the least cost by an optimal transport assignment method [18,25,26,98].”
- DETRs Beat YOLOs on Real-time Object Detection (2023, methodology): “YOLO detectors can be classified into two categories: anchor-based [1, 10, 14, 24, 27, 28, 34, 35] and anchor-free [9, 11, 15, 37], which achieve a reasonable trade-off between speed and accuracy and are widely used in various practical scenarios.”
- CBNet: A Composite Backbone Network Architecture for Object Detection (2021, methodology): “As shown in Table VIII, our CBCSPNet-L improves CSPNet-L [84] by 2.”
- HunyuanVideo: A Systematic Framework For Large Video Generative Models (2024, positive): “We also develop YOLOX [24]-like visual models to detect and remove some occluded or sensitive information such as watermarks, borders, and logos.”
- A review of convolutional neural network architectures and their optimizations (2022, positive): “2% compared to the end-to-end approach (Ge et al. 2021).”
- Simple Multi-dataset Detection (2021, background): “This fragmentation brought rapid progress in object detection [5, 10, 20, 31, 39, 45] and instance segmentation [14], but comes with a drawback.”
- Hyper-YOLO: When Visual Object Detection Meets Hypergraph Computation (2024, neutral): “YOLOX [9] introduced anchor-free detection, and Gold-YOLO [10] enhanced feature fusion with its Gather-and-Distribute mechanism.”
- AlphaPose: Whole-Body Regional Multi-Person Pose Estimation and Tracking in Real-Time (2022, methodology): “For the detectionmodule, we adopt YOLOX [81] YOLOV3-SPP [64], EfficientDet [75] and JDE [67].”
- A Comprehensive Review of YOLO: From YOLOv1 to YOLOv8 and Beyond (2023, methodology): “Following the trends of the time driven by the works of [81, 80, 79, 78], PP-YOLOE uses an anchor-free architecture.”
- A review of convolutional neural networks in computer vision (2024, positive): “The anchor-free paradigm is employed as the training approach, and to further increase the detection accuracy, the SimOTA (Ge et al. 2021) label assignment technique and SIoU (Gevorgyan 2022) bounding box regression loss are included.”
- GDRNPP: A Geometry-Guided and Fully Learning-Based Object Pose Estimator (2021, neutral): “In the core, we first detect all objects of interest usinganoff-the-shelfobjectdetector,suchas[80],[81],[82].”
- YOLOv9: Learning What You Want to Learn Using Programmable Gradient Information (2024, methodology): “The current mainstream real-time object detectors are the YOLO series [2, 7, 13–15, 25, 30, 31, 47–49, 61–63, 74, 75], and most of these models use CSPNet [64] or ELAN [65] and their variants as the main computing units.”
- PP-YOLOE: An evolved version of YOLO (2022, methodology): “Except for the stem, the depth setting of basic backbone is [3, 6, 6, 3].”