Articles

Apprentissage profond en reconnaissance d'images: Techniques et défis

L’apprentissage profond en reconnaissance d’images est une technologie puissante qui permet de résoudre des problèmes complexes. Découvrez les techniques et les défis associés à cette technologie.

Dans le vaste royaume de l’intelligence artificielle, l’apprentissage profond est devenu un jeu-changer, en particulier dans le domaine de la reconnaissance d’images. La capacité des machines à reconnaître et à catégoriser des images, à la manière du cerveau humain, a ouvert une multitude d’opportunités et de défis. Plongeons-nous dans les techniques que l’apprentissage profond offre pour la reconnaissance d’images et les obstacles qui y sont associés.

Data: For CNNs to work, large amounts of data are required. The more data that is available, the more accurate the results will be. This is because the network needs to be trained on a variety of images, so it can learn to recognize patterns and distinguish between different objects.

Hurdles: The main challenge with CNNs is that they require a lot of data and computing power. This can be expensive and time-consuming, and it can also lead to overfitting if not enough data is available. Additionally, CNNs are not able to generalize well, meaning they are not able to recognize objects that they have not been trained on.

Réseaux de neurones convolutionnels (CNN)

Technique : Les CNN sont le pilier des systèmes de reconnaissance d’images modernes. Ils se composent de plusieurs couches de petites collections de neurones qui traitent des parties de l’image d’entrée, appelées champs réceptifs. Les résultats de ces collections sont ensuite assemblés de manière à se chevaucher, afin d’obtenir une meilleure représentation de l’image d’origine ; c’est une caractéristique distinctive des CNN.

Données : Pour que les CNN fonctionnent, des quantités importantes de données sont nécessaires. Plus il y a de données disponibles, plus les résultats seront précis. C’est parce que le réseau doit être formé sur une variété d’images, afin qu’il puisse apprendre à reconnaître des modèles et à distinguer différents objets.

Hurdles : Le principal défi avec les CNN est qu’ils nécessitent beaucoup de données et de puissance de calcul. Cela peut être coûteux et prendre du temps, et cela peut également entraîner un surajustement si pas assez de données sont disponibles. De plus, les CNN ne sont pas en mesure de généraliser bien, ce qui signifie qu’ils ne sont pas en mesure de reconnaître des objets qu’ils n’ont pas été formés.

Réseaux neuronaux profonds (DNN)

Technique : Les DNN sont une variante des CNN qui peuvent être utilisés pour la reconnaissance d’images. Ils sont constitués de plusieurs couches de neurones qui traitent des parties de l’image d’entrée et produisent des résultats plus précis que les CNN. Les DNN peuvent également être utilisés pour la classification d’images et la segmentation d’images.

Données : Les DNN nécessitent également des grandes quantités de données pour fonctionner correctement. Cependant, ils peuvent être entraînés sur des jeux de données plus petits que les CNN et peuvent donc être plus efficaces lorsqu’il n’y a pas assez de données disponibles.

Hurdles : Le principal défi avec les DNN est qu’ils nécessitent beaucoup de temps et de puissance de calcul pour être entraînés correctement. De plus, ils sont sensibles aux bruit et aux variations dans les données d’entrée, ce qui peut entraîner des résultats imprécis.

Source de l’article sur DZONE

This article is an excerpt from the book Machine Learning with PyTorch and Scikit-Learn from the best-selling Python Machine Learning series, updated and expanded to cover PyTorch, transformers, and graph neural networks.

Broadly speaking, graphs represent a certain way we describe and capture relationships in data. Graphs are a particular kind of data structure that is nonlinear and abstract. And since graphs are abstract objects, a concrete representation needs to be defined so the graphs can be operated on. Furthermore, graphs can be defined to have certain properties that may require different representations. Figure 1 summarizes the common types of graphs, which we will discuss in more detail in the following subsections:
Common types of graph

Source de l’article sur DZONE

In the wake of COVID, video streaming is no longer a fun diversion. Organizations are depending on it to keep their workforce moving… and parents are counting on it to keep their kids from going into all-out rebellion mode during lockdowns. We’re all familiar with the hiccups we experience using streaming platforms, so the application of Deep Learning to video encoding and streaming promises to be an interesting frontier. 

Convolutional Neural Networks (CNNs) are a form of Deep Learning – machine learning designed to mimic the human brain by creating multiple layers of ‘neuron’ connections based on weighted probabilities – that is commonly used in image recognition. Each neuron represents a combination of features from a dataset, which are activated for prediction through sigmoid, threshold and rectifier functions. 

Source de l’article sur DZONE

The real value of a modern DataOps platform is realized only when business users and applications are able to access raw and aggregated data from a range of sources, and produce data-driven insights in a timely manner. And with Machine Learning (ML), analysts and data scientists can leverage historical data to help make better, data-driven business decisions-offline and in real-time using technologies such as TensorFlow.

In this post, you will learn how to use TensorFlow (TF) models for prediction and classification using the newly released TensorFlow Evaluator* in StreamSets Data Collector 3.5.0 and StreamSets Data Collector Edge.

Source de l’article sur DZONE


The AI vs. ML Dilemma

Machine Learning and Artificial Intelligence are two terms that are used interchangeably all the time. So, are AI and ML the same? Most definitely not!

Any technology that makes a system exhibit human-like intelligence is AI. Machine Learning is actually one type of AI. Machine Learning makes decisions by relying on the use of mathematical models that are trained on data. ML models are capable of making better decisions when more data is available.


Source de l’article sur DZONE (AI)

You must have seen videos on Youtube or posts on your news feed in which certain texts or a person’s face is blurred. Well, that’s how our digital privacy is ensured by simplest of technologies.
But think about it, in an age of Machine Learning, can’t your digital privacy be easily breached? The answer is a big "Yes," and a team of researchers at the University of Texas has proven that. They have developed a software that can identify the sensitive content hidden behind blurred or pixelated images. The content can be someone’s house or vehicle number, or simply a human face.

Interestingly, the team hasn’t used some state of the art technology to do it. It has instead used Machine Learning methods to train the neural networks. So instead of being programmed, the computer has been fed with large volumes of sample images. The algorithm used doesn’t actually unblur or restore the image. It identifies the content of the blurred image based on the information it already has.


Source de l’article sur DZONE (AI)

Machine vision, or computer vision, is a popular research topic in artificial intelligence (AI) that has been around for many years. However, machine vision still remains as one of the biggest challenges in AI. In this article, we will explore the use of deep neural networks to address some of the fundamental challenges of computer vision. In particular, we will be looking at applications such as network compression, fine-grained image classification, captioning, texture synthesis, image search, and object tracking.

Network Compression

Even though deep neural networks feature incredible performance, their demands for computing power and storage pose a significant challenge to their deployment in actual application. Research shows that the parameters used in a neural network can be hugely redundant. Therefore, a lot of work is put into increasing accuracy while also decreasing the complexity of the network.


Source de l’article sur DZONE (AI)

Image titleComputers today can not only automatically classify photos, but they can also describe the various elements in pictures and write short sentences describing each segment with proper English grammar. This is done by the Deep Learning Network (CNN), which actually learns patterns that naturally occur in photos. Imagenet is one of the biggest databases of labeled images to train the Convolutional Neural Networks using GPU-accelerated Deep Learning frameworks such as Caffe2, Chainer, Microsoft Cognitive Toolkit, MXNet, PaddlePaddle, Pytorch, TensorFlow, and inference optimizers such as TensorRT.

Neural Networks were first used in 2009 for speech recognition and were only implemented by Google in 2012. Deep Learning, also called Neural Networks, is a subset of Machine Learning that uses a model of computing that’s very much inspired by the structure of the brain.


Source de l’article sur DZONE (AI)

If you are planning to experiment with deep learning models, Keras might be a good place to start. It’s a high-level API written in Python with backend support for Tensorflow, CNTK, and Theano.

For those of you who are new to Keras, you can read more at keras.io or a simple google search will take you to the basics and more on Keras.


Source de l’article sur DZONE (AI)

Only a year ago, industry discourse around artificial intelligence (AI) was focused on whether or not to go the AI way. Businesses found themselves facing an important choice — weighing the considerable value that would manifest against the investment of capital and talent AI would necessitate. But that was yesterday.

Today, we have reached a critical inflection point. With their technology deployments hitting maturity, early adopters of AI have begun to realize incredible advantages — the ability to optimize operations, maximize productivity, derive insights and be more responsive to real-time market demands. The results are out for the world to see.


Source de l’article sur DZONE (AI)