Recognizing Unseen Objects via Multimodal Intensive Knowledge Graph Propagation
3
Citation
35
Reference
10
Related Paper
Citation Trend
Abstract:
Zero-Shot Learning (ZSL), which aims at automatically recognizing unseen objects, is a promising learning paradigm to understand new real-world knowledge for machines continuously. Recently, the Knowledge Graph (KG) has been proven as an effective scheme for handling the zero-shot task with large-scale and non-attribute data. Prior studies always embed relationships of seen and unseen objects into visual information from existing knowledge graphs to promote the cognitive ability of the unseen data. Actually, real-world knowledge is naturally formed by multimodal facts. Compared with ordinary structural knowledge from a graph perspective, multimodal KG can provide cognitive systems with fine-grained knowledge. For example, the text description and visual content can depict more critical details of a fact than only depending on knowledge triplets. Unfortunately, this multimodal fine-grained knowledge is largely unexploited due to the bottleneck of feature alignment between different modalities. To that end, we propose a multimodal intensive ZSL framework that matches regions of images with corresponding semantic embeddings via a designed dense attention module and self-calibration loss. It makes the semantic transfer process of our ZSL framework learns more differentiated knowledge between entities. Our model also gets rid of the performance limitation of only using rough global features. We conduct extensive experiments and evaluate our model on large-scale real-world data. The experimental results clearly demonstrate the effectiveness of the proposed model in standard zero-shot classification tasks.Keywords:
Knowledge graph
Visual reasoning
Bottleneck detection in manufacturing is the key to improving production efficiency and stability in order to improve capacity. Yet common bottleneck detection methods in industry and academia lack either accuracy or practicability, or both, for dynamic systems. The new methodology is conducted by the observation of processes and inventories. Blocked processes and full inventories indicate a downstream bottleneck. Starved processes and empty inventories indicate an upstream bottleneck. Through subsequent observations of multiple process states and inventory levels within a system, it is possible to determine the direction of the bottleneck at the given time and hence to find the momentary bottleneck in the system. The shifting of bottlenecks can be observed directly. Work-sampling techniques can be used to obtain a long-term picture of the dynamically shifting bottleneck. The new methodology does not require any calculations, statistics, or time measurements. Hence the method is suited for practical use by shop floor supervisors and clerks. The direct observation of the bottleneck also gives additional information about the underlying causes of the bottlenecks, simplifying the improvement of the system capacity. Extensive field testing of the method received positive feedback not only from management but also shop floor operators. The method is already in use at the Robert Bosch GmbH, where it is known as the bottleneck walk.
Cite
Citations (32)
Bottleneck shiftiness is an important managerial problem that negatively affects shop floor manageability. It has therefore received much research attention. Yet research has focused on how protective capacity can be used to influence bottleneck shiftiness rather than on assessing its operational impact. The latter is complex to evaluate since changing the degree of bottleneck shiftiness influences utilization, which makes the results of different experimental settings non-comparable. To overcome this problem, we take a different approach. Bottleneck shiftiness is decomposed by investigating its underlying phenomenon: the impact of the bottleneck position. Using simulation, we demonstrate that tighter control can be exercised, and better performance achieved, the further upstream the bottleneck is positioned. It is consequently important to be aware of the direction of the bottleneck shift. If the bottleneck shifts upstream, performance is likely to improve rather than deteriorate as is implicitly assumed in the literature.
Position (finance)
Cite
Citations (23)
Identification
Traffic system
Cite
Citations (11)
A variety of established approaches exist for the detection of dynamic bottlenecks. Furthermore, the prediction of bottlenecks is experiencing a growing scientific interest, quantifiable by the increasing number of publications in recent years. Neglected, on the other hand, is the diagnosis of occurring bottlenecks. Detection methods may determine the current location of a bottleneck, while predictive approaches may indicate the location of an upcoming bottleneck. However, mere knowledge of current and future bottlenecks does not enable concrete actions to be taken to avoid the bottlenecks, nor does it open up any immediate advantage for manufacturing companies. Since small and medium-sized companies in particular have limited resources, they cannot implement improvement measures for every bottleneck that occurs. Due to the shifts of dynamic bottlenecks, the selection of the mostsuitable stations in the value stream becomes more difficult. This paper therefore contributes to the neglected field of bottleneck diagnosis. First, we propose two data-driven metrics, relative bottleneck frequency and relative bottleneck severity, which allow a quantitative assessment of the respective bottleneck situations. For validation purposes, we apply these metrics in nine selected scenarios generated using discrete event simulation in a value stream with a serial manufacturing line. Finally, we evaluate and discuss the results.
Cite
Citations (0)
The performance of a manufacturing or logistic system is determined by its constraints. Therefore, in order to improve the performance, it is necessary to improve the constraints, also known as the bottlenecks. Finding the bottlenecks, however, is not easy. We compare the two most common bottleneck detection methods, based on the utilization and the waiting time, with the shifting bottleneck detection method developed by us, for AGV systems. We find that the two conventional methods have many shortcomings compared to the shifting bottleneck detection method. In the example presented here, conventional methods are either unable to detect the bottleneck at all or detect the bottleneck incorrectly. The shifting bottleneck detection method not only finds the bottlenecks but also determines the magnitude of the primary and secondary bottlenecks.
Cite
Citations (54)
Buffer (optical fiber)
Cite
Citations (11)
Cite
Citations (3)
Theory of constraints (TOC) claims that each production system has at least one bottleneck resource and the bottleneck determines throughput of the whole system. Now more and more managers regard to exploit bottleneck resources. However, both literatures and facts show that bottleneck shiftiness does exit and sometimes is unavoidable. It makes bottleneck management problems become more complicated. This paper studies some factors that often cause bottleneck shift from one machine to another, including product-mix, production lot-size and load-balanced level, and puts forward corresponding solutions to overcome these problems based on basic principles of TOC, for providing enterprises some useful advices to prevent the appearance of those unexpected bottlenecks, or to reduce the unexpected production fluctuate when the shifting is unavoidable
Theory of Constraints
Cite
Citations (5)
The performance of a manufacturing or logistic system is determined by its constraints. Therefore, in order to improve the performance, it is necessary to improve the constraints, also known as the bottlenecks. Finding the bottlenecks, however, is not easy. This paper compares the two most common bottleneck detection methods, based on the utilization and the waiting time, with the shifting bottleneck detection method developed by us, for AGV systems. We find that the two conventional methods have many shortcomings compared to the shifting bottleneck detection method. In the example presented here, conventional methods are either unable to detect the bottleneck at all or detect the bottleneck incorrectly. The shifting bottleneck detection method not only finds the bottlenecks but also determines the magnitude of the primary and secondary bottlenecks.
Cite
Citations (10)
As a weak point, bottleneck plays a crucial role in improving the property of the whole water system. In this paper, sensitivity computation formula is defined and corresponding mathematical model is established to confirm the bottleneck of water network. And the case proves that the bottleneck streams and units can be easily confirmed by the method in this paper.
Cite
Citations (0)