Bottleneck Block Research Articles

Hierarchical Binary CNNs for Landmark Localization with Limited Resources

Our goal is to design architectures that retain the groundbreaking performance of Convolutional Neural Networks (CNNs) for landmark localization and at the same time are lightweight, compact and suitable for applications with limited computational resources. To this end, we make the following contributions: (a) we are the first to study the effect of neural network binarization on localization tasks, namely human pose estimation and face alignment. We exhaustively evaluate various design choices, identify performance bottlenecks, and more importantly propose multiple orthogonal ways to boost performance. (b) Based on our analysis, we propose a novel hierarchical, parallel and multi-scale residual architecture that yields large performance improvement over the standard bottleneck block while having the same number of parameters, thus bridging the gap between the original network and its binarized counterpart. (c) We perform a large number of ablation studies that shed light on the properties and the performance of the proposed block. (d) We present results for experiments on the most challenging datasets for human pose estimation and face alignment, reporting in many cases state-of-the-art performance. (e) We further provide additional results for the problem of facial part segmentation. Code can be downloaded from https://www.adrianbulat.com/binary-cnn-landmarks.

An Integrated Process Planning and Scheduling System for Block Assembly in Shipbuilding

This paper describes an integrated process planning and scheduling system for block assembly in shipbuilding. The system consists of process planning module, scheduling module, bottleneck block selection module, and process re-planning module. A rule-based reasoning technology is applied to determine optimal assembly units and assembly sequences in generating initial process plans. For scheduling module, a schedule revision heuristic is developed for efficient reallocation of blocks to alternative assembly shops. For bottleneck block selection which plays a central role in bridging process planning and scheduling, a powerful heuristic is developed by employing an entropy-based partitioning method for identifying bottleneck periods. The experimental results with data from a real shipyard show that our search heuristics are very effective and efficient.