Hit-Detector: Hierarchical Trinity Architecture Search for Object Detection

Guo, Jianyuan; Han, Kai; Wang, Yunhe; Zhang, Chao; Yang, Zhaohui; Wu, Han; Chen, Xinghao; Xu, Chang

doi:10.1109/cvpr42600.2020.01142

Cited by 92 publications

(59 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…anchor matching and refinement [30,31], and enrichment of features [32,33]. One stage detector uses the anchor as a reference box for final selections.…”

Section: A Anchor-based Object Detectionmentioning

confidence: 99%

Multi-Level Refinement Feature Pyramid Network for Scale Imbalance Object Detection

Aziz

Sheikh

et al. 2021

IEEE Access

View full text Add to dashboard Cite

“…anchor matching and refinement [30,31], and enrichment of features [32,33]. One stage detector uses the anchor as a reference box for final selections.…”

Section: A Anchor-based Object Detectionmentioning

confidence: 99%

Multi-Level Refinement Feature Pyramid Network for Scale Imbalance Object Detection

Aziz

Sheikh

et al. 2021

IEEE Access

View full text Add to dashboard Cite

“…The target architecture is derived according to the distribution of architectural parameters. Due to its high efficiency, many works extend DNAS to more applications, including semantic segmentation [15,25], object detection [9,10], etc. Some DNAS works [5,8,22] propose to integrate co-optimization with both accuracy and hardware properties into the search phase.…”

Section: Related Workmentioning

confidence: 99%

Exploring Complicated Search Spaces with Interleaving-Free Sampling

Tian¹,

Xie²,

Fang³

et al. 2021

Preprint

View full text Add to dashboard Cite

The existing neural architecture search algorithms are mostly working on search spaces with short-distance connections. We argue that such designs, though safe and stable, obstacles the search algorithms from exploring more complicated scenarios. In this paper, we build the search algorithm upon a complicated search space with long-distance connections, and show that existing weightsharing search algorithms mostly fail due to the existence of interleaved connections. Based on the observation, we present a simple yet effective algorithm named IF-NAS, where we perform a periodic sampling strategy to construct different sub-networks during the search procedure, avoiding the interleaved connections to emerge in any of them. In the proposed search space, IF-NAS outperform both random sampling and previous weight-sharing search algorithms by a significant margin. IF-NAS also generalizes to the micro cell-based spaces which are much easier. Our research emphasizes the importance of macro structure and we look forward to further efforts along this direction.

show abstract

“…Although there have been huge improvements for accurate object detection [21], the struggle for high-speed and energy-efficient detectors is largely unsolved. Several recent approaches exploit neural architecture search (NAS) to discover efficient and accurate object detectors [3,13,11]. On the meanwhile, there are some attempts to utilize model compression methods for improving the efficiency of detectors, e.g.,channel pruning [1], quantization [43] and knowledge distillation [4,42].…”

Section: Related Workmentioning

confidence: 99%

“…Another family of solutions is to use common model compression methods for accelerating object detectors, e.g., knowledge distillation [4,42] and pruning [1]. Moreover, some recent works utilize neural architecture search (NAS) approach for searching better architectures for different components of object detectors [3,13,11]. Although these methods mentioned above show strong performance while improving the efficiency, they are mainly built with traditional convolutional neural networks, which contain massive inefficient multiplications.…”

Section: Introductionmentioning

confidence: 99%

An Empirical Study of Adder Neural Networks for Object Detection

Chen

Dong

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

Adder neural networks (AdderNets) have shown impressive performance on image classification with only addition operations, which are more energy efficient than traditional convolutional neural networks built with multiplications. Compared with classification, there is a strong demand on reducing the energy consumption of modern object detectors via AdderNets for real-world applications such as autonomous driving and face detection. In this paper, we present an empirical study of AdderNets for object detection. We first reveal that the batch normalization statistics in the pre-trained adder backbone should not be frozen, since the relatively large feature variance of AdderNets. Moreover, we insert more shortcut connections in the neck part and design a new feature fusion architecture for avoiding the sparse features of adder layers. We present extensive ablation studies to explore several design choices of adder detectors. Comparisons with state-of-the-arts are conducted on COCO and PASCAL VOC benchmarks. Specifically, the proposed Adder FCOS achieves a 37.8% AP on the COCO val set, demonstrating comparable performance to that of the convolutional counterpart with an about 1.4× energy reduction.

show abstract

Hit-Detector: Hierarchical Trinity Architecture Search for Object Detection

Cited by 92 publications

References 29 publications

Multi-Level Refinement Feature Pyramid Network for Scale Imbalance Object Detection

Multi-Level Refinement Feature Pyramid Network for Scale Imbalance Object Detection

Exploring Complicated Search Spaces with Interleaving-Free Sampling

An Empirical Study of Adder Neural Networks for Object Detection

Contact Info

Product

Resources

About