Pelee: A Real-Time Object Detection System on Mobile Devices

By Robert J. Wang, Xiang Li, Shuang Ao, Charles X. F. Ling. 2018

517 citations recorded by the source corpus.

An increasing need of running Convolutional Neural Network (CNN) models on mobile devices with limited computing power and memory resource encourages studies on efficient model design. A number of efficient architectures have been proposed in recent years, for example, MobileNet, ShuffleNet, and MobileNetV2. However, all these models are heavily dependent on depthwise separable convolution which lacks efficient implementation in most deep learning frameworks. In this study, we propose an efficient architecture named PeleeNet, which is built with conventional convolution instead. On ImageNet ILSVRC 2012 dataset, our proposed PeleeNet achieves a higher accuracy and over 1.8 times faster speed than MobileNet and MobileNetV2 on NVIDIA TX2. Meanwhile, PeleeNet is only 66% of the model size of MobileNet. We then propose a real-time object detection system by combining PeleeNet with Single Shot MultiBox Detector (SSD) method and optimizing the architecture for fast speed. Our proposed detection system2, named Pelee, achieves 76.4% mAP (mean average precision) on PASCAL VOC2007 and 22.4 mAP on MS COCO dataset at the speed of 23.6 FPS on iPhone 8 and 125 FPS on NVIDIA TX2. The result on COCO outperforms YOLOv2 in consideration of a higher precision, 13.6 times lower computational cost and 11.3 times smaller model size.

Subfields: Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

Citations in the corpus

Fast Neural Network Adaptation via Parameter Remapping and Architecture Search (2020, background): “PeleeNet (Wang et al., 2018) proposes a variant of DenseNet (Huang et al., 2017) for more real-time object detection on mobile devices.”
PC-DARTS: Partial Channel Connections for Memory-Efficient Architecture Search (2019, neutral): “5 Pelee (Wang et al., 2018) 304×304 PeleeNet 1.”
Data-Driven Sparse Structure Selection for Deep Neural Networks (2017, methodology): “To prove the practicability of our method, we further experiment in a very lightweight network, PeleeNet [42].”
MiniVLM: A Smaller and Faster Vision-Language Model (2020, neutral): “[37, 48, 52, 33, 20].”
ThunderNet: Towards Real-Time Generic Object Detection on Mobile Devices (2019, methodology): “Prior lightweight one-stage detectors [11, 28, 31, 13] do not obtain an ideal accuracy/speed trade-off: there is a huge accuracy gap between them and the large detectors [19, 25], while they fail to achieve real-time detection on mobile devices.”
A Survey of Modern Deep Learning based Object Detection Models (2021, methodology): “PeleeNet was centered around the DenseNet [23] but looked at many other models for inspiration.”
Mish: A Self Regularized Non-Monotonic Activation Function (2020, methodology): “3, we compare the Top-1 and Top-5 accuracy of Mish against ReLU, Leaky ReLU, and Swish on PeleeNet [46] , Cross Stage Partial ResNet-50 [45], and ResNet-18/50 [15].”
A Survey of Deep Learning-Based Object Detection (2019, methodology): “[24] propose a novel real-time object detection system by combining PeleeNet with SSD [10] and optimizing the architecture for fast processing speed.”
CSPNet: A New Backbone that can Enhance Learning Capability of CNN (2019, methodology): “For real-time object detection on CPU or mobile GPU, SSD-based Pelee [37], YOLOv3-based PRN [35], and Light-Head RCNN [17]-based ThunderNet [25] all receive excellent performance on object detection.”
BiSeNet V2: Bilateral Network with Guided Aggregation for Real-Time Semantic Segmentation (2020, methodology): “The ﬁrst group is non-real-time methods, containing CRF-RNN (Zheng et al., 2015), Deeplab-CRF (Chen et al., 2015), FCN-8S (Long et al., 2015), Dilation10 (Yu and Koltun, 2016), LRR (Ghiasi and Fowlkes, 2016), Deeplabv2-CRF (Chen et al., 2016), FRRN (Pohlen et al., 2017), ReﬁneNet (Lin et al., 2017),”
Object Detection in 20 Years: A Survey (2019, unknown): “In recent years, the bottle-neck design has been widely used for designing lightweight networks [50], [133], [146], [147], [148].”
CornerNet-Lite: Efficient Keypoint based Object Detection (2019, background): “PeleeNet [55], in contrast, demonstrates the effectiveness of standard convolu-tions by introducing an efﬁcient variant of DenseNet [17] consisting of two-way dense layers and a stem block.”
Distilling Object Detectors via Decoupled Features (2021, background): “Techniques such as quantization [19, 58, 31, 57, 62], pruning [2, 17, 20], network design [55, 49, 15, 18] and knowledge distillation [56, 6] have been developed to overcome this dilemma and achieve an efficient inference on detection task.”
Review of Image Classification Algorithms Based on Convolutional Neural Networks (2021, unknown): “In 2018, PeleeNet [168] makes improvements based on DenseNet [131], which is an efficient architecture for embedded platforms.”
Partial Order Pruning: For Best Speed/Accuracy Trade-Off in Neural Architecture Search (2019, background): “This also indicates the FLOPs may be inconsistent with latency on the target platform [27, 20].”