YOLOv7: Trainable Bag-of-Freebies Sets New State-of-the-Art for Real-Time Object Detectors
By Chien-Yao Wang, Alexey Bochkovskiy, Hongpeng Liao. Computer Vision and Pattern Recognition, 2022
10,740 citations recorded by the source corpus.
Real-time object detection is one of the most important research topics in computer vision. As new approaches regarding architecture optimization and training optimization are continually being developed, we have found two research topics that have spawned when dealing with these latest state-of-the-art methods. To address the topics, we propose a trainable bag-of-freebies oriented solution. We combine the flexible and efficient training tools with the proposed architecture and the compound scaling method. YOLOv7 surpasses all known object detectors in both speed and accuracy in the range from 5 FPS to 120 FPS and has the highest accuracy 56.8% AP among all known real-time object detectors with 30 FPS or higher on GPU V100. Source code is released in https://github.com/WongKinYiu/yolov7.
Subfields: Aerospace Engineering, Computer Vision and Pattern Recognition
View Paper: https://www.semanticscholar.org/paper/3aed4648f7857c1d5e9b1da4c3afaf97463138c3
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Aerospace Engineering
- Muhammad Hussain
- Diana-Margarita Córdova-Esparza
- Juan R. Terven
- I-Hau Yeh
- Hao Zhang
- Shuaijie Zhang
- Cong Xu
- Hongpeng Liao
In Computer Vision and Pattern Recognition
- Dacheng Tao
- Luc Van Gool
- Hongpeng Liao
- Chien-Yao Wang
- Xinggang Wang
- Ping Luo
- Chunhua Shen
- Xiaogang Wang
Citations in the corpus
- Inner-IoU: More Effective Intersection over Union Loss with Auxiliary Bounding Box (2023, methodology): “Anchor based algorithms include Faster R-CNN [14], YOLO (You Only Look Once) series [10]–[13], SSD (Single Shot MultiBox Detector) [15] and RetinaNet [16] Anchor free detection algorithms include CornerNet [17], CenterNet [18] and FCOS (Fully Convolutional One Stage Object Detection) [19].”
- YOLOv10: Real-Time End-to-End Object Detection (2024, methodology): “YOLOv7 [62] introduces E-ELAN for rich gradient flow path and explores several trainable bag-of-freebies methods.”
- Slim-neck by GSConv: a lightweight-design for real-time detector architectures (2022, methodology): “The Yolo family detectors [4-7, 44, 46] are widely used in industry because of the efficient detection.”
- FFCA-YOLO for Small Object Detection in Remote Sensing Images (2024, background): “YOLO series of algorithm [18], [19], [20], as typical one-stage object detection algorithms, has advantages to achieve desired performance for small objects.”
- A Comprehensive Review of YOLO: From YOLOv1 to YOLOv8 and Beyond (2023, result): “YOLOv7 [94] was published in ArXiv in July 2022 by the same authors of YOLOv4 and YOLOR.”
- DETRs Beat YOLOs on Real-time Object Detection (2023, methodology): “We utilize the benchmark to test the end-to-end speed of anchor-based detectors YOLOv5 [10] and YOLOv7 [35], as well as anchor-free detectors PP-YOLOE [37], YOLOv6 [15] and YOLOv8 [11] on T4 GPU with TensorRT FP16.”
- Gold-YOLO: Efficient Object Detector via Gather-and-Distribute Mechanism (2023, methodology): “To compare our Gold-YOLO with other state-of-the-art detectors in the YOLO series, such as YOLOv5 [13], YOLOX [11], PPYOLOE [52], YOLOv7 [48], YOLOv8 [14] and YOLOv6-3.”
- ASF-YOLO: A Novel YOLO Model with Attentional Scale Sequence Fusion for Cell Instance Segmentation (2023, methodology): “YOLOMask [31], PR-YOLO [32] and YOLO-SF [33] augmented YOLOv5 [8] and YOLOv7-Tiny [34] with Convolutional Block Attention Module (CBAM) [35].”
- YOLOv9: Learning What You Want to Learn Using Programmable Gradient Information (2024, methodology): “All experimental setups follow YOLOv7 AF [63], while the dataset is MS COCO 2017 splitting.”
- A Comprehensive Review of YOLO Architectures in Computer Vision: From YOLOv1 to YOLOv8 and YOLO-NAS (2023, result): “YOLOv7 [106] was published in ArXiv in July 2022 by the same authors of YOLOv4 and YOLOR.”
- YOLOv1 to v8: Unveiling Each Variant–A Comprehensive Review of YOLO (2024, methodology): “The architectural footprint, incorporated an efficient backbone featuring RepVGG/CSPStackRep blocks, a PAN (Path Aggregation Network) topology neck, and an efficient decoupled head with a hybrid-channel strategy [43].”
- Hyper-YOLO: When Visual Object Detection Meets Hypergraph Computation (2024, positive): “This architecture synergistically blends three typical convolutional variants: the 1 × 1 bypass convolution for channel-wise feature recalibration, the Depthwise Separable Convolution (DSConv) for efficient spatial feature processing, and the C2f module for enhancing feature hierarchy integration.”
- YOLO-World: Real-Time Open-Vocabulary Object Detection (2024, methodology): “Several works [11, 23, 33, 52, 55] propose various architectures or designs for YOLO, including path aggregation networks [29], cross-stage partial networks [51], and re-parameterization [7], which further improve both speed and accuracy.”
- YOLOv6: A Single-Stage Object Detection Framework for Industrial Applications (2022, methodology): “We notice that a half-stride gray border is put around each image when evaluating the model performance in the implementations of YOLOv5 [10] and YOLOv7 [42].”
- UAV-YOLOv8: A Small-Object-Detection Model Based on Improved YOLOv8 for UAV Aerial Photography Scenarios (2023, methodology): “One-stage detectors include methods such as the you only look once (YOLO) series [8–13] and single-shot multibox detector (SSD) [14], which use a separate neural network for one forward inference to generate the coordinates and category results of the prediction boxes.”
- Brain Tumor Detection Based on Deep Learning Approaches and Magnetic Resonance Imaging (2023, background): “Network Architechture of YOLOv7 Outstanding results on publicly available datasets have led to YOLOv7′s widespread recognition as a state-of-the-art object detector [70].”
- YOLO-based Object Detection Models: A Review and its Applications (2024, neutral): “After a month of release of YOLOv6, the YOLO v7 [88] was released by the same author of YOLOv4.”
- 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].”
- AFPN: Asymptotic Feature Pyramid Network for Object Detection (2023, methodology): “One-stage methods [1]–[3] predict the category and location of objects directly from the input image.”