XNOR-Net: ImageNet Classification Using Binary Convolutional Neural Networks

Abstract

We propose two efficient approximations to standard convolutional neural networks: Binary-Weight-Networks and XNOR-Networks. In Binary-Weight-Networks, the filters are approximated with binary values resulting in 32x memory saving. In XNOR-Networks, both the filters and the input to convolutional layers are binary. XNOR-Networks approximate convolutions using primarily binary operations. This results in 58x faster convolutional operations and 32x memory savings. XNOR-Nets offer the possibility of running state-of-the-art networks on CPUs (rather than GPUs) in real-time. Our binary networks are simple, accurate, efficient, and work on challenging visual tasks. We evaluate our approach on the ImageNet classification task. The classification accuracy with a Binary-Weight-Network version of AlexNet is only 2.9% less than the full-precision AlexNet (in top-1 measure). We compare our method with recent network binarization methods, BinaryConnect and BinaryNets, and outperform these methods by large margins on ImageNet, more than 16% in top-1 accuracy.

References in corpus (8)

• Very Deep Convolutional Networks for Large-Scale Image Recognition
• SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and <0.5MB model size
• Deep Residual Learning for Image Recognition
• Compressing Deep Convolutional Networks using Vector Quantization
• Compressing Neural Networks with the Hashing Trick
• Speeding up Convolutional Neural Networks with Low Rank Expansions
• Bitwise Neural Networks
• Big Neural Networks Waste Capacity

Cited by in corpus (82)

• MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications
• DoReFa-Net: Training Low Bitwidth Convolutional Neural Networks with Low Bitwidth Gradients
• Quantized Neural Networks: Training Neural Networks with Low Precision Weights and Activations
• Ternary Weight Networks
• Trained Ternary Quantization
• Pruning Filters for Efficient ConvNets
• Demystifying Parallel and Distributed Deep Learning: An In-Depth Concurrency Analysis
• What-and-Where to Match: Deep Spatially Multiplicative Integration Networks for Person Re-identification
• DeepCABAC: A Universal Compression Algorithm for Deep Neural Networks
• Ristretto: Hardware-Oriented Approximation of Convolutional Neural Networks
• Deep Convolutional Neural Network Inference with Floating-point Weights and Fixed-point Activations
• Embedded Binarized Neural Networks
• Deep neural networks are robust to weight binarization and other non-linear distortions
• Distributed Deep Neural Networks over the Cloud, the Edge and End Devices
• NSGA-Net: Neural Architecture Search using Multi-Objective Genetic Algorithm
• Effective Quantization Methods for Recurrent Neural Networks
• SLSNet: Skin lesion segmentation using a lightweight generative adversarial network
• A neural network memory prefetcher using semantic locality
• Towards the Limit of Network Quantization
• Trained Quantization Thresholds for Accurate and Efficient Fixed-Point Inference of Deep Neural Networks
• Compacting Deep Neural Networks for Internet of Things: Methods and Applications
• ShiftCNN: Generalized Low-Precision Architecture for Inference of Convolutional Neural Networks
• A Survey on Deep Learning Methods for Robot Vision
• Delta Networks for Optimized Recurrent Network Computation
• Searching for Low-Bit Weights in Quantized Neural Networks
• Machine Learning Models that Remember Too Much
• FINN-R: An End-to-End Deep-Learning Framework for Fast Exploration of Quantized Neural Networks
• Streamlined Deployment for Quantized Neural Networks
• Training DNNs with Hybrid Block Floating Point
• Exploration of Low Numeric Precision Deep Learning Inference Using Intel FPGAs
• Not All Ops Are Created Equal!
• ShaResNet: reducing residual network parameter number by sharing weights
• Binarized Neural Networks on the ImageNet Classification Task
• Accelerating Deep Convolutional Networks using low-precision and sparsity
• Quantization and Training of Low Bit-Width Convolutional Neural Networks for Object Detection
• An Experimental Study of the Impact of Pre-training on the Pruning of a Convolutional Neural Network
• A Configurable BNN ASIC using a Network of Programmable Threshold Logic Standard Cells
• Local Feature Detectors, Descriptors, and Image Representations: A Survey
• Towards Evolutional Compression
• Toolflows for Mapping Convolutional Neural Networks on FPGAs: A Survey and Future Directions
• PAMS: Quantized Super-Resolution via Parameterized Max Scale
• Efficient Stochastic Inference of Bitwise Deep Neural Networks
• Recent Advances in the Applications of Convolutional Neural Networks to Medical Image Contour Detection
• Structured Convolution Matrices for Energy-efficient Deep learning
• Full deep neural network training on a pruned weight budget
• BinaryRelax: A Relaxation Approach For Training Deep Neural Networks With Quantized Weights
• HEMP: High-order Entropy Minimization for neural network comPression
• Training Bit Fully Convolutional Network for Fast Semantic Segmentation
• Optimizing for Interpretability in Deep Neural Networks with Tree Regularization
• Structured Deep Neural Network Pruning via Matrix Pivoting
• Efficient Hybrid Network Architectures for Extremely Quantized Neural Networks Enabling Intelligence at the Edge
• Accelerating Binarized Neural Networks via Bit-Tensor-Cores in Turing GPUs
• Fractional Skipping: Towards Finer-Grained Dynamic CNN Inference
• Automating Generation of Low Precision Deep Learning Operators
• Convolutional Neural Network Simplification with Progressive Retraining
• HarDNet: A Low Memory Traffic Network
• CNN Acceleration by Low-rank Approximation with Quantized Factors
• Low-Power Computer Vision: Status, Challenges, Opportunities
• A GPU-Outperforming FPGA Accelerator Architecture for Binary Convolutional Neural Networks
• Binarized Convolutional Neural Networks with Separable Filters for Efficient Hardware Acceleration
• A High-Performance Adaptive Quantization Approach for Edge CNN Applications
• DeepLight: Deep Lightweight Feature Interactions for Accelerating CTR Predictions in Ad Serving
• The Challenge of Multi-Operand Adders in CNNs on FPGAs: How not to solve it!
• Towards Fast and Energy-Efficient Binarized Neural Network Inference on FPGA
• Fast Video Classification via Adaptive Cascading of Deep Models
• On Psychoacoustically Weighted Cost Functions Towards Resource-Efficient Deep Neural Networks for Speech Denoising
• Always-On 674uW @ 4GOP/s Error Resilient Binary Neural Networks with Aggressive SRAM Voltage Scaling on a 22nm IoT End-Node
• DNN Feature Map Compression using Learned Representation over GF(2)
• Understanding the Energy and Precision Requirements for Online Learning
• Quantized neural network design under weight capacity constraint
• Neural network compression via learnable wavelet transforms
• Coarse and fine-grained automatic cropping deep convolutional neural network
• WeClick: Weakly-Supervised Video Semantic Segmentation with Click Annotations
• In-memory Implementation of On-chip Trainable and Scalable ANN for AI/ML Applications
• Scientific Calculator for Designing Trojan Detectors in Neural Networks
• Multi-Precision Quantized Neural Networks via Encoding Decomposition of -1 and +1
• Auto Deep Compression by Reinforcement Learning Based Actor-Critic Structure
• AE-Netv2: Optimization of Image Fusion Efficiency and Network Architecture
• clcNet: Improving the Efficiency of Convolutional Neural Network using Channel Local Convolutions
• Reliable Identification of Redundant Kernels for Convolutional Neural Network Compression
• Prune the Convolutional Neural Networks with Sparse Shrink
• Modulated binary cliquenet