SqueezeDet: Unified, Small, Low Power Fully Convolutional Neural Networks for Real-Time Object Detection for Autonomous Driving

By Bichen Wu, Forrest N. Iandola, Peter H. Jin, Kurt Keutzer. 2016

539 citations recorded by the source corpus.

Object detection is a crucial task for autonomous driving. In addition to requiring high accuracy to ensure safety, object detection for autonomous driving also requires realtime inference speed to guarantee prompt vehicle control, as well as small model size and energy efficiency to enable embedded system deployment.,,,,,, In this work, we propose SqueezeDet, a fully convolutional neural network for object detection that aims to simultaneously satisfy all of the above constraints. In our network we use convolutional layers not only to extract feature maps, but also as the output layer to compute bounding boxes and class probabilities. The detection pipeline of our model only contains a single forward pass of a neural network, thus it is extremely fast. Our model is fullyconvolutional, which leads to small model size and better energy efficiency. Finally, our experiments show that our model is very accurate, achieving state-of-the-art accuracy on the KITTI [10] benchmark. The source code of SqueezeDet is open-source released.

Subfields: Automotive Engineering, 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 Automotive Engineering

In Computer Vision and Pattern Recognition

Citations in the corpus

Delay-Aware Microservice Coordination in Mobile Edge Computing: A Reinforcement Learning Approach (2019, background): “1, where an autonomous vehicle detects cars, pedestrians, cyclists, road signs, and other objects in real time to make the appropriate control decisions for driving safety [21].”
Improving Multispectral Pedestrian Detection by Addressing Modality Imbalance Problems (2020, background): “dim environment, face spoofing detection [47], autonomous driving [24,39], etc), the introducing of another modality dramatically benefits the tasks of object detection.”
A Robust Real-Time Automatic License Plate Recognition Based on the YOLO Detector (2018, methodology): “As great advances in object detection were achieved through YOLO-inspired models [9], [10], we decided to fine-tune it”
DAWNBench : An End-to-End Deep Learning Benchmark and Competition (2017, unknown): “Researchers have proposed new software systems [29, 1, 8, 12, 9, 46], training algorithms [33, 48, 45, 28, 44, 18, 27, 50, 42, 43, 14], communication methods [39, 49, 11, 9, 12, 22] and hardware [20, 30, 37, 19, 7, 21] to decrease this cost.”
A crossbar array of magnetoresistive memory devices for in-memory computing (2022, methodology): “We demonstrate our MRAM array for this face detection application by using it as the seventh (convolution) layer in a ten-layer VGG-like model modified from SqueezeDet 38 , with all the rest layers computed in software.”
Real-Time Object Detection With Reduced Region Proposal Network via Multi-Feature Concatenation (2020, background): “2929059 pedestrian detection [6], have emerged, and object detection has gained significance [7].”
Model Compression and Acceleration for Deep Neural Networks: The Principles, Progress, and Challenges (2018, methodology): “SqueezeNet [48] was proposed to replace 3 × 3 convolution with 1 × 1 convolution, which created a compact neural network with”
Deep Multi-Modal Object Detection and Semantic Segmentation for Autonomous Driving: Datasets, Methods, and Challenges (2019, background): “as PASCAL VOC [19] and COCO [20], and have been widely applied in autonomous driving, including detecting traffic lights [21]–[24], road signs [25]–[28], people [29]–[32], or vehicles [33]–[38], to name a few.”
A Survey of Model Compression and Acceleration for Deep Neural Networks (2017, methodology): “Baseline Models Representative Works Alexnet [1] structural matrix [30]–[32] low-rank factorization [39] Network in network [68] low-rank factorization [39] VGG nets [69] transfered filters [43] low-rank factorization [39] Residual networks [70] compact filters [48], stochastic depth [61] parameter ”
Knowledge-based Recurrent Attentive Neural Network for Traffic Sign Detection (2018, methodology): “And then our proposed KB-RANN (a) Faster RCNN (b) SqueezeDet (c) KB-RANN (Our Method) algorithm gets the data of detection result after image preprocessing.”
Scenic: a language for scenario specification and scene generation (2018, methodology): “Specifically, we replaced a random 5% of X matrix (250 images) with images from X overlap , keeping the overall training set size constant.”
Gaussian YOLOv3: An Accurate and Fast Object Detector Using Localization Uncertainty for Autonomous Driving (2019, methodology): “Because there is no GT for testing, the training and validation sets are made by randomly splitting the training set in half [25].”
ExpandNets: Linear Over-parameterization to Train Compact Convolutional Networks (2018, background): “In a parallel line of research, several works have proposed design strategies to reduce a network’s number of parameters [21, 36, 43, 48, 53, 58].”