An Energy and GPU-Computation Efficient Backbone Network for Real-Time Object Detection

By Youngwan Lee, Joong-won Hwang, Sangrok Lee, Yuseok Bae, Jongyoul Park. 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), 2019

502 citations recorded by the source corpus.

As DenseNet conserves intermediate features with diverse receptive fields by aggregating them with dense connection, it shows good performance on the object detection task. Although feature reuse enables DenseNet to produce strong features with a small number of model parameters and FLOPs, the detector with DenseNet backbone shows rather slow speed and low energy efficiency. We find the linearly increasing input channel by dense connection leads to heavy memory access cost, which causes computation overhead and more energy consumption. To solve the inefficiency of DenseNet, we propose an energy and computation efficient architecture called VoVNet comprised of One-Shot Aggregation (OSA). The OSA not only adopts the strength of DenseNet that represents diversified features with multi receptive fields but also overcomes the inefficiency of dense connection by aggregating all features only once in the last feature maps. To validate the effectiveness of VoVNet as a backbone network, we design both lightweight and large-scale VoVNet and apply them to one-stage and two-stage object detectors. Our VoVNet based detectors outperform DenseNet based ones with 2× faster speed and the energy consumptions are reduced by 1.6× - 4.1×. In addition to DenseNet, VoVNet also outperforms widely used ResNet backbone with faster speed and better energy efficiency. In particular, the small object detection performance has been significantly improved over DenseNet and ResNet.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Neurology

View Paper: https://www.semanticscholar.org/paper/b517f179ffd609247f0665e7247fbebb7a4e8ea3

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

In Neurology

Citations in the corpus

Designing Network Design Strategies Through Gradient Path Analysis (2022, methodology): “ELAN is mainly composed of VoVNet [17] combined with CSPNet [33], and optimizes the gradient length of the overall network with the structure of stack in computational block .”
Exploring Object-Centric Temporal Modeling for Efficient Multi-View 3D Object Detection (2023, methodology): “To scale up our method, we also report results on the nuScenes test set with V2-99 initialized from DD3D [38] checkpoint and ViT-Large [7].”
End-to-End Autonomous Driving: Challenges and Frontiers (2023, background): “Depth-pre-trained CNNs [176] signiﬁcantly boost perception and downstream performance.”
MPViT: Multi-Path Vision Transformer for Dense Prediction (2021, methodology): “Modern CNN backbones which show better performance for dense prediction leverage multiple scales at the convolutional kernel level [19, 32, 33, 46, 47], or feature level [34, 42, 55].”
PP-YOLOE: An evolved version of YOLO (2022, methodology): “VoVNet[15] and subsequent TreeNet[20] also show superior performance in object detection and instance segmentation.”
BEVFormer: Learning Bird's-Eye-View Representation from Multi-Camera Images via Spatiotemporal Transformers (2022, methodology): “Following previous methods [45, 47, 31], we adopt two types of backbone: ResNet101-DCN [15, 12] that initialized from FCOS3D [45] checkpoint, and VoVnet-99 [21] that initialized from DD3D [31] checkpoint.”
Hyper-YOLO: When Visual Object Detection Meets Hypergraph Computation (2024, unknown): “Unlike representation learning in computer vision only processes visual features, those hypergraph computation meth-ods [18], [32] simultaneously process features and high-order structures.”
YOLOv7: Trainable Bag-of-Freebies Sets New State-of-the-Art for Real-Time Object Detectors (2022, methodology): “The design of CSPVoVNet [79] in Figure 2 (b) is a variation of VoVNet [39].”
Review of Image Classification Algorithms Based on Convolutional Neural Networks (2021, methodology): “[172] proposed VoVNetV1 that the One-shot aggregation (OSA) module is designed which is more efficient than Dense Block [131].”
Scaled-YOLOv4: Scaling Cross Stage Partial Network (2020, methodology): “We will use the CSPOSANet with PCB architecture to form the backbone of YOLOv4.”