NAS-FCOS: Fast Neural Architecture Search for Object Detection

By Ning Wang, Yang Gao, Hao Chen, Peng Wang, Zhi Tian, Chunhua Shen. 2019

229 citations recorded by the source corpus.

The success of deep neural networks relies on significant architecture engineering. Recently neural architecture search (NAS) has emerged as a promise to greatly reduce manual effort in network design by automatically searching for optimal architectures, although typically such algorithms need an excessive amount of computational resources, e.g., a few thousand GPU-days. To date, on challenging vision tasks such as object detection, NAS, especially fast versions of NAS, is less studied. Here we propose to search for the decoder structure of object detectors with search efficiency being taken into consideration. To be more specific, we aim to efficiently search for the feature pyramid network (FPN) as well as the prediction head of a simple anchor-free object detector, namely FCOS, using a tailored reinforcement learning paradigm. With carefully designed search space, search algorithms and strategies for evaluating network quality, we are able to efficiently search a top-performing detection architecture within 4 days using 8 V100 GPUs. The discovered architecture surpasses state-of-the-art object detection models (such as Faster R-CNN, RetinaNet and FCOS) by 1.5 to 3.5 points in AP on the COCO dataset, with comparable computation complexity and memory footprint, demonstrating the efficacy of the proposed NAS for object detection.

Subfields: Artificial Intelligence, 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 Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

Memory-Efficient Hierarchical Neural Architecture Search for Image Denoising (2019, methodology): “An alternative to EA is to use reinforcement learning (RL) techniques, e.g., policy gradients [47, 36] and Q-learning [44], to train a recurrent neural network that acts as a meta-controller to generate potential architectures—typically encoded as sequences—by exploring a predeﬁned search space.”
CBNet: A Composite Backbone Network Architecture for Object Detection (2021, background): “8% AP, outperforming previous anchor-free methods [7], [30], [31], [32], [36], [37], [64] under single-scale testing protocol.”
Auto-Panoptic: Cooperative Multi-Component Architecture Search for Panoptic Segmentation (2020, unknown): “In this paper, inspired by the advances achieved by Network Architecture Search (NAS) on fundamental vision tasks [27, 2, 36, 41], we aim to explore an efﬁcient foreground-background cooperative architecture automatically tailored for panoptic segmentation, namely Auto-Panoptic.”
Retinex-inspired Unrolling with Cooperative Prior Architecture Search for Low-light Image Enhancement (2020, background): “The above search algorithms have achieved highly competitive performance in various high-level vision tasks, including image classiﬁcation [36], object detection [30], and semantic segmentation.”
RTMDet: An Empirical Study of Designing Real-Time Object Detectors (2022, methodology): “To improve the model efficiency, efficient backbone networks and model scaling strategies [3, 41, 71] and enhancement of multi-scale feature [7, 23, 36, 46, 49, 68] are explored either by handcrafted design or neural architecture search [10, 17, 23, 74].”
SM-NAS: Structural-to-Modular Neural Architecture Search for Object Detection (2019, methodology): “NAS-FPN[14], Auto-FPN[58], NAS-FCOS[53] use NAS to find a better feature fusion neck and a more powerful RCNN head.”
Dilated SpineNet for Semantic Segmentation (2021, background): “Recent architectures discovered by NAS have shown promising results than handcrafted models on vision tasks including image classification [57, 31, 41, 20], object detection [16, 14, 13, 23, 43, 46, 24], semantic segmentation [27, 34], etc.”
DetectoRS: Detecting Objects with Recursive Feature Pyramid and Switchable Atrous Convolution (2020, unknown): “, [40, 49, 54, 60, 64,73,86,87], and multi-stage methods, e.”
Seesaw Loss for Long-Tailed Instance Segmentation (2020, background): “Classical single-stage approaches [28, 31] require densely populated anchors as a prior, while anchor-free methods [24, 25, 45, 51] manage to achieve similar or better performance without such prior.”