Spatially Adaptive Computation Time for Residual Networks
arXiv:1612.02297
Abstract
This paper proposes a deep learning architecture based on Residual Network that dynamically adjusts the number of executed layers for the regions of the image. This architecture is end-to-end trainable, deterministic and problem-agnostic. It is therefore applicable without any modifications to a wide range of computer vision problems such as image classification, object detection and image segmentation. We present experimental results showing that this model improves the computational efficiency of Residual Networks on the challenging ImageNet classification and COCO object detection datasets. Additionally, we evaluate the computation time maps on the visual saliency dataset cat2000 and find that they correlate surprisingly well with human eye fixation positions.
CVPR 2017
References in corpus (11)
- Very Deep Convolutional Networks for Large-Scale Image Recognition
- R-FCN: Object Detection via Region-based Fully Convolutional Networks
- Densely Connected Convolutional Networks
- Going Deeper with Convolutions
- Recurrent Models of Visual Attention
- Multiple Object Recognition with Visual Attention
- Multimodal Compact Bilinear Pooling for Visual Question Answering and Visual Grounding
- Bridging the Gaps Between Residual Learning, Recurrent Neural Networks and Visual Cortex
- Highway and Residual Networks learn Unrolled Iterative Estimation
- VideoLSTM Convolves, Attends and Flows for Action Recognition
- BranchyNet: Fast Inference via Early Exiting from Deep Neural Networks
Cited by in corpus (16)
- CondenseNet: An Efficient DenseNet using Learned Group Convolutions
- Distributed Deep Neural Networks over the Cloud, the Edge and End Devices
- SpotTune: Transfer Learning through Adaptive Fine-tuning
- BlockDrop: Dynamic Inference Paths in Residual Networks
- AMPNet: Asynchronous Model-Parallel Training for Dynamic Neural Networks
- Dynamically Sacrificing Accuracy for Reduced Computation: Cascaded Inference Based on Softmax Confidence
- IamNN: Iterative and Adaptive Mobile Neural Network for Efficient Image Classification
- Improved Techniques for Training Adaptive Deep Networks
- Dynamic Computational Time for Visual Attention
- BlockCopy: High-Resolution Video Processing with Block-Sparse Feature Propagation and Online Policies
- SGAD: Soft-Guided Adaptively-Dropped Neural Network
- Stochastic Downsampling for Cost-Adjustable Inference and Improved Regularization in Convolutional Networks
- Early Improving Recurrent Elastic Highway Network
- Energy-efficient Amortized Inference with Cascaded Deep Classifiers
- BABO: Background Activation Black-Out for Efficient Object Detection
- Sparsely ensembled convolutional neural network classifiers via reinforcement learning