BlockDrop: Dynamic Inference Paths in Residual Networks

• Published in: 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition pp. 8817 - 8826
• Main Authors: Wu, Zuxuan, Nagarajan, Tushar, Kumar, Abhishek, Rennie, Steven, Davis, Larry S., Grauman, Kristen, Feris, Rogerio
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.06.2018
• Subjects: Computational modeling; Computer vision; Dogs; Neural networks; Task analysis; Training; Visualization

Very deep convolutional neural networks offer excellent recognition results, yet their computational expense limits their impact for many real-world applications. We introduce BlockDrop, an approach that learns to dynamically choose which layers of a deep network to execute during inference so as to...