Deep neuralnetworks have been successfully deployed in various domains of artificial intelligence, including computer vision and natural language processing. We observe that the current standard procedure for trainin...
详细信息
ISBN:
(纸本)9781577358664
Deep neuralnetworks have been successfully deployed in various domains of artificial intelligence, including computer vision and natural language processing. We observe that the current standard procedure for training DNNs discards all the learned information in the past epochs except the current learned weights. An interesting question is: is this discarded information indeed useless? We argue that the discarded information can benefit the subsequent training. In this paper, we propose learning with retrospection (LWR) which makes use of the learned information in the past epochs to guide the subsequent training. LWR is a simple yet effective training framework to improve accuracies, calibration, and robustness of DNNs without introducing any additional network parameters or inference cost, but only with a negligible training overhead. Extensive experiments on several benchmark datasets demonstrate the superiority of LWR for training DNNs.
The early detection of diseases and pests in coffee crops by means of artificial vision and pattern recognition brings with it the ease of inspection and the reduction of crop losses in coffee plantations. This work p...
详细信息
ISBN:
(纸本)9781665498432
The early detection of diseases and pests in coffee crops by means of artificial vision and pattern recognition brings with it the ease of inspection and the reduction of crop losses in coffee plantations. This work proposes a model based on neuralnetworks that is capable of detecting coffee leaves in an image and also classifies them into the most common diseases in the Panamanian tropics. The coffee leaf disease classification model is able to classify the following diseases: Cercospora leaf spot, leaf rust, leaf miner and phoma. This model obtained an accuracy of 100% and a loss of 1.6 x 10(-5) during the training phase. Subsequent to the training phase, a validation of the model was performed;during this phase, an overall accuracy of 90% was obtained for each of the diseases and pests. In future work, it is desired to implement this architecture in the vision system of the agricultural robot. The vision system will allow farmers to effectively inspect and manage their crops.
Gaussian noise is one of the main noise sources of digital images. In this paper, the idea of judging before processing is adopted to improve the efficiency of Gaussian noise removal. First, the Pascal VOC dataset is ...
详细信息
ISBN:
(数字)9798350373820
ISBN:
(纸本)9798350373837
Gaussian noise is one of the main noise sources of digital images. In this paper, the idea of judging before processing is adopted to improve the efficiency of Gaussian noise removal. First, the Pascal VOC dataset is taken as a clean image, and the Gaussian noise image dataset is obtained by adding Gaussian noise with random density. Then, various conventional and lightweight convolutional neural network models are constructed, and the constructed datasets are imported for training, validating and testing. Experimental results show that Vgg-16 achieves the highest classification accuracy of 0.9929. Due to its high computational complexity, VGG-16 is only applicable to scenarios that do not require mobile applications such as servers and hosts. SqueezeNet achieves the second-highest classification accuracy of 0.9906 with relatively low computational complexity, making it suitable for portable consumer electronics such as smartphones and smartwatches. Gaussian noise image recognition is realized based on conventional and lightweight convolutional neural network models, which provides a technical basis for efficient Gaussian denoising.
Cell Spheroids are of high interest for clinical cell applications and cell screening. To allow the extraction of early readout parameters a high amount of image data of petri dishes is created. To support automated a...
详细信息
Functional connectivity plays an essential role in modern neuroscience. The modality sheds light on the brain's functional and structural aspects, including mechanisms behind multiple pathologies. One such patholo...
详细信息
ISBN:
(数字)9781665490627
ISBN:
(纸本)9781665490627
Functional connectivity plays an essential role in modern neuroscience. The modality sheds light on the brain's functional and structural aspects, including mechanisms behind multiple pathologies. One such pathology is schizophrenia which is often followed by auditory verbal hallucinations. The latter is commonly studied by observing functional connectivity during speech processing. In this work, we have made a step toward an in-depth examination of functional connectivity during a dichotic listening task via deep learning for three groups of people: schizophrenia patients with and without auditory verbal hallucinations and healthy controls. We propose a graph neural network-based framework within which we represent EEG data as signals in the graph domain. The framework allows one to 1) predict a brain mental disorder based on EEG recording, 2) differentiate the listening state from the resting state for each group and 3) recognize characteristic task-depending connectivity. Experimental results show that the proposed model can differentiate between the above groups with state-of-the-art performance. Besides, it provides a researcher with meaningful information regarding each group's functional connectivity, which we validated on the current domain knowledge.
Literacy is a fundamental and intrinsic right for every person;therefore, government agencies, NGOs, and private donors seek to implement sustainable and long-term plans that allow citizens to access quality, inclusiv...
详细信息
ISBN:
(纸本)9783030991708;9783030991692
Literacy is a fundamental and intrinsic right for every person;therefore, government agencies, NGOs, and private donors seek to implement sustainable and long-term plans that allow citizens to access quality, inclusive, and equitable education. But despite their efforts, illiteracy has not yet been fully eradicated, so educational programs and tools have been created to work to solve this problem. In this research, a web application is developed that takes the resources of the Google QuickDraw API to generate an interactive image interpretation tool, which aims to support the learning of reading and writing in people with illiteracy. The tool allows an illiterate person to draw pictograms that are interpreted by the software and through convolutional neuralnetworks with deep learning, a prediction of the image is generated. The result is presented in a graphical user interface, which displays a real image corresponding to the prediction together with text and audio assistance, essential for user learning. The results show a performance of 75% accuracy in the training and 56% in the test. The performance tests yielded indicate that the tool meets the minimum software requirements for its future implementation in institutions that work in the education of people with illiteracy.
In the era of pervasive Internet use, managing large volumes of image data becomes crucial. To mitigate storage and bandwidth costs, image compression plays a pivotal role. Traditional image compression techniques lik...
详细信息
ISBN:
(数字)9798331506520
ISBN:
(纸本)9798331506537
In the era of pervasive Internet use, managing large volumes of image data becomes crucial. To mitigate storage and bandwidth costs, image compression plays a pivotal role. Traditional image compression techniques like JPEG and PNG, while widely used, may suffer from limited compression rates. In this work, we propose a novel approach using shallow Convolutional neural Network (CNN) autoencoders for on-the-fly image compression. Our model aims to achieve high compression rates with improved image quality compared to classical methods. Additionally, it supports decompression on the CPU in realtime, making no assumptions about client-side computational resources. We present a comprehensive methodology, including architecture design, experiments, and performance metrics. Our results demonstrate the effectiveness of the proposed approach, providing a competitive alternative for online content compression and decompression that outperforms JPEG in image quality assessment metrics at 33% compression rate. We also outperform other Learned image Compression (LIC) techniques in both the decompression time and number of trainable parameters with an improvement in the order of 10-times.
Plastic waste has become a significant environmental concern, necessitating innovative approaches to enhance recycling processes. Traditional methods of sorting plastic waste are labor intensive and error prone, leadi...
详细信息
BP neural network is using gradient descent method to continuously adjust the weights and thresholds between the input layer and the hidden layer, so that the error decreases along the direction of the gradient, and t...
详细信息
ISBN:
(数字)9798350360240
ISBN:
(纸本)9798350384161
BP neural network is using gradient descent method to continuously adjust the weights and thresholds between the input layer and the hidden layer, so that the error decreases along the direction of the gradient, and the training stops when the variance between the actual output layer and the desired output layer reaches the minimum, and obtains the threshold of the best training performance. However, the drawbacks of the gradient descent method are reflected in the control training of the steady pendulum motion of the cart based on BP neural network and gradient descent method, that is, although the direction of the gradient descent is the direction of the fastest reduction of the loss function, this is not necessarily the direction of the fastest convergence. And at the same time, since the gradient descent method utilizes the local nature of the objective function, it is very likely to fall into the local minimum in the iteration. To address the above drawbacks, we compare the training results using gradient descent, Newton's method, and conjugate gradient method, and find that Newton's method as well as conjugate gradient method have faster convergence, higher recognition rate, and better training results.
With the development of artificial Intelligence (AI), deep neuralnetworks have been used in a variety of fields such as digital signal processing, object detection as well as medical use. It acts as a vital aid to pe...
详细信息
暂无评论