EfficientDet: Scalable and Efficient Object Detection
By Mingxing Tan, Ruoming Pang, Quoc V. Le. 2019
7,168 citations recorded by the source corpus.
Model efficiency has become increasingly important in computer vision. In this paper, we systematically study neural network architecture design choices for object detection and propose several key optimizations to improve efficiency. First, we propose a weighted bi-directional feature pyramid network (BiFPN), which allows easy and fast multi-scale feature fusion; Second, we propose a compound scaling method that uniformly scales the resolution, depth, and width for all backbone, feature network, and box/class prediction networks at the same time. Based on these optimizations and EfficientNet backbones, we have developed a new family of object detectors, called EfficientDet, which consistently achieve much better efficiency than prior art across a wide spectrum of resource constraints. In particular, with single-model and single-scale, our EfficientDetD7 achieves state-of-the-art 52.2 AP on COCO test-dev with 52M parameters and 325B FLOPs, being 4x – 9x smaller and using 13x – 42x fewer FLOPs than previous detector.
Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Efficiency & Compression
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
- Jianfeng Gao
- Quoc V. Le
- Mohit Bansal
- Yoshua Bengio
- Sergey Levine
- Dacheng Tao
- Alan L. Yuille
- Percy Liang
In Computer Vision and Pattern Recognition
- Luc Van Gool
- Dacheng Tao
- Xiaogang Wang
- Ming-Hsuan Yang
- Alan L. Yuille
- Trevor Darrell
- Jitendra Malik
- Chunhua Shen
In Efficiency & Compression
- Quoc V. Le
- Tijmen Blankevoort
- Diganta Misra
- Markus Nagel
- Song Han
- Kurt Keutzer
- Mingxing Tan
- Yunhe Wang
Citations in the corpus
- Combined Scaling for Zero-shot Transfer Learning (2021, methodology): “The second dimension is the model size, with representative works such as EfficientNet, VITs and related works (Tan and Le, 2019, 2021; Tan et al., 2020; Dosovitskiy et al., 2021; Zhai et al., 2021; Bello et al., 2021).”
- Masked-attention Mask Transformer for Universal Image Segmentation (2021, methodology): “For instance, although multiple approaches have been proposed for building feature pyramid representations [33], as we show in our experiments, BiFPN [47] performs better for instance segmentation while FaPN [39] performs better for semantic segmentation.”
- Towards Real-World Prohibited Item Detection: A Large-Scale X-ray Benchmark (2021, methodology): “BiFPN [33] is weighted bi-directional feature pyramid network, which allows easy and fast multi-scale feature fusion.”
- Simple Copy-Paste is a Strong Data Augmentation Method for Instance Segmentation (2020, background): “3 — — EfficientDet-D7x (1536) [57] 410B 77M 54.”
- Deformable DETR: Deformable Transformers for End-to-End Object Detection (2020, background): “2 EfficientDet-D7 (Tan et al., 2020) EfficientNet-B6 52.”
- MaX-DeepLab: End-to-End Panoptic Segmentation with Mask Transformers (2020, methodology): “This stacked decoder is similar to feature pyramid networks [58, 63, 84, 76] designed for pyramidal anchor predictions [64], but our purpose here is only to aggregate multi-scale features, i.”
- AutoML: A Survey of the State-of-the-Art (2019, methodology): “Therefore, NAS has also been applied to many other computer vision tasks, such as object detection [153, 154, 155], semantic segmentation [156, 157], image restoration [158], GAN [159], graph neural network (GNN) [160].”
- AutoHAS: Differentiable Hyper-parameter and Architecture Search (2020, background): “2% Neural Architecture Search (NAS) has brought significant improvements in many applications, such as machine perception [19, 48, 5, 51, 50], language modeling [32, 11], and model compression [17, 10, 16].”
- HiFi-GAN: Generative Adversarial Networks for Efficient and High Fidelity Speech Synthesis (2020, methodology): “The synthesis speed was measured on GPU and CPU environments according to the recent research trends regarding the efficiency of neural networks (Kumar et al., 2019, Zhai et al., 2020, Tan et al., 2020).”
- MDETR - Modulated Detection for End-to-End Multi-Modal Understanding (2021, methodology): “In our work, we are not limited by the existence of pre-trained detectors, and inspired by its success in object detection [58], we choose to explore the EfficientNet family [57] for our backbone.”
- Bottleneck Transformers for Visual Recognition (2021, methodology): “Using BoTNet, we demonstrate significantly improved results on instance segmentation without any bells and whistles such as Cascade R-CNN [7], FPN changes [41, 19, 43, 57], hyperparameter changes [56], etc.”
- Dynamic Head: Unifying Object Detection Heads with Attentions (2021, methodology): “Our method is competitive and more efficient to learn compared with EffcientDet [27] and SpineNet [8], with a significantly less 1 / 20 training time.”
- Learning Data Augmentation Strategies for Object Detection (2019, background): “Much work in object detection was devoted to building better model architectures or detection algorithms [11, 35, 22, 13, 10, 39, 34, 33, 32].”
- EfficientNetV2: Smaller Models and Faster Training (2021, methodology): “, 2018), object detection (Chen et al., 2019; Tan et al., 2020), segmentation (Liu et al.”
- YOLOv4: Optimal Speed and Accuracy of Object Detection (2020, methodology): “Networks equipped with this mechanism include Feature Pyramid Network (FPN) [44], Path Aggregation Network (PAN) [49], BiFPN [77], and NAS-FPN [17].”
- Compressing Visual-linguistic Model via Knowledge Distillation (2021, methodology): “Following MiniVLM [71], we also adopt the EfficientNet [66] based lightweight object detector (TEE) for visual feature extraction.”
- You Only Look One-level Feature (2021, methodology): “line of studies of designing complex fusion methods manually [22, 14, 25], or via Neural Architecture Search (NAS) algorithms [7, 34].”