Litcius/Paper detail

Single-Shot Image Recognition Using Siamese Neural Networks

Abhiraj Malhotra

202311 citationsDOI

Abstract

Generating useful characteristics for machine learning algorithms may be quite pricey operationally & can be challenging in some cases instances wherein there is a lack of information. The one-shot learn context, in that we should accurately anticipate provided just an unique sample of every fresh category, is a classic illustration of it. In this study, we investigate a technique for training Siamese neural nets, which use a special framework to prioritize input similarities. When one net has been trained, we may next take advantage of strong exclusionary characteristics to extend the program's prediction capacity to completely fresh categories with uncertain probabilities as well as fresh information. With a convolutional structure, researchers are capable to provide robust outcomes that are superior to similar deep learning systems that operate almost at the cutting edge on one-off categorization problems.

Topics & Concepts

CategorizationComputer scienceConvolutional neural networkArtificial intelligenceEnhanced Data Rates for GSM EvolutionMachine learningContext (archaeology)Shot (pellet)Artificial neural networkSample (material)One shotImage (mathematics)Pattern recognition (psychology)Deep learningChemistryMechanical engineeringPaleontologyChromatographyBiologyEngineeringOrganic chemistryHuman Pose and Action RecognitionGenerative Adversarial Networks and Image SynthesisAnomaly Detection Techniques and Applications