Researchers introduce an Interaction‑aware Spatio‑Temporal Pyramid Attention network that embeds a PCA‑guided loss to capture complementary multi‑scale features, enabling end‑to‑end video action classification with state‑of‑the‑art accuracy on UCF101, HMDB51, Charades and internal datasets.
Recent advances in CNN‑based single image super‑resolution assume bicubic down‑sampling, limiting real‑world performance; this paper introduces a dimension‑stretching strategy that feeds blur kernels and noise levels into a CNN, enabling a single model (SRMD) to efficiently handle multiple, even spatially varying, degradation types with strong quantitative and visual results.
This tutorial walks through rewriting the HED edge‑detection network for TensorFlow 1.7, covering deprecated API fixes, migration from TF‑Slim to tf.layers, matrix initialization, batch normalization nuances, and a comprehensive review of convolution variants such as 1×1, depthwise, separable, and dilated convolutions, plus guidance on transposed convolutions and modern architectures like ResNet and Inception.
This article explains how to implement client‑side human voice recognition on mobile devices using TensorFlow Lite, detailing the mel‑spectrogram feature extraction, algorithmic optimizations such as ARM instruction set and multithreading, model selection with Inception‑v3 CNN, training procedures, and deployment steps.
This article introduces the fundamentals and evolution of optical flow, covering classic algorithms such as Horn‑Schunck and Lucas‑Kanade, modern deep‑learning approaches like FlowNet, and their practical applications in video detection, semantic segmentation, and novel view synthesis.
This article explains the evolution of optical character recognition, outlines the complete OCR processing pipeline—including image input, preprocessing, binarization, noise removal, tilt correction, layout analysis, character segmentation, recognition, and post‑processing—while showcasing Ctrip's real‑world OCR project, its architecture, accuracy metrics, and key computer‑vision techniques such as CNN, HSV, HOG, LBP, and Haar features.
This study compares hash‑based, CNN‑based, and local‑feature methods for client‑side image similarity detection in e‑commerce, showing that while hash methods are fast and CNNs are accurate but costly, the Hessian‑Affine detector combined with SIFT descriptors delivers the optimal balance of computational efficiency, robustness to transformations, and high recall/precision for on‑device duplicate filtering.
This guide walks you through installing Anaconda, creating a TensorFlow virtual environment on Windows, configuring CPU and GPU versions, and implementing both a basic softmax regression and a deep convolutional neural network for MNIST digit recognition, complete with code snippets, training tips, and visualization tools.
This beginner‑friendly TensorFlow tutorial by Chen Yidong walks readers through Windows environment setup, explains TensorFlow’s graph‑execution model, and demonstrates both softmax linear regression and a deep convolutional neural network for MNIST, while also covering utility scripts, TensorBoard visualization, and CPU/GPU or multi‑GPU deployment.
This article provides a comprehensive guide to text classification, covering traditional pipelines, bag‑of‑words and TF‑IDF features, dimensionality‑reduction techniques, word‑embedding models such as GloVe, word2vec and fastText, and modern deep‑learning architectures like CNN, RCNN and HAN.
This article explains why machine learning is essential for complex tasks, defines neural networks, outlines three reasons to study them, and provides concise overviews of eight fundamental neural network architectures—including perceptron, CNN, RNN, LSTM, Hopfield, Boltzmann machines, deep belief networks, and deep autoencoders—grouped by their structural categories.
The WhereAreYou hackathon project demonstrates an iOS AR app that visualizes nearby ride‑hailing cars by converting multiple CNN models to Core ML, using ARKit to map GPS bearings, then switching to vision‑based YOLO‑style detection and tracking for real‑time vehicle identification and distance labeling.
This article details how 58.com leverages deep learning models such as FNN, Wide&Deep, CNN+DNN, and YouTube DNN recall, along with a custom AI platform, to enhance recommendation ranking and recall, achieving measurable improvements in click‑through rates and overall system performance.
This article explains the motivation, core principles, and implementation details of Batch Normalization, including how it normalizes each mini‑batch, restores learned feature distributions, and is applied in convolutional neural networks to speed up training and boost model generalization.
This article surveys the evolution of face recognition from traditional image‑based methods to real‑time video processing, highlights key researchers and open‑source projects, explains the four‑stage pipeline, details DeepFace's deep‑learning architecture, and provides practical installation and usage instructions for Python developers.
This article provides a comprehensive introduction to machine learning and deep learning, covering their definitions, classifications, key algorithms, neural network structures, core concepts such as generalization and regularization, and typical architectures like CNN and RNN, illustrated with numerous diagrams.
This article explains the challenges of image‑based porn detection, details the multi‑label classification approach of the Wukong system, and reveals the deep‑learning techniques—including CNN evolution, transfer learning, loss functions, adversarial training, and GAN‑based data augmentation—that enable over 99.5% accuracy with massive daily request volumes.
This article surveys recent breakthroughs in deep learning for computer vision, explains what constitutes an advanced architecture, outlines common vision tasks, and provides concise overviews plus paper and Keras implementation links for ten influential models such as AlexNet, VGG, ResNet, and GAN.
This article explores how deep learning, specifically convolutional neural networks built with Caffe, can be used to automatically detect and predict static UI compatibility bugs such as text overlap in mobile app front‑ends, reducing reliance on manual testing.
This article presents a deep‑learning approach for recognizing seven basic human facial expressions using a balanced FER2013 dataset, describes the CNN architecture built with Keras and OpenCV preprocessing, reports training on AWS GPU, and analyzes validation results and visualizations.
This article walks through a real‑world mobile implementation of a convolutional neural network for document detection, covering problem definition, limitations of traditional OpenCV pipelines, the adoption of a HED edge‑detection network, data preparation, model training, TensorFlow library trimming, and deployment tricks for iOS and Android.
Recent advances in image super-resolution leverage deep learning techniques such as convolutional neural networks, residual learning, perceptual loss, generative adversarial networks, and PixelCNN to reconstruct high-resolution details from low-resolution inputs, addressing challenges of scalability, training efficiency, and multi-scale upscaling.
This article introduces the fundamentals of neural networks for image recognition, explains regression vs classification, describes convolution, pooling and fully connected layers, illustrates the classic LeNet‑5 model on the MNIST dataset, and shows how a TensorFlow‑based CNN can be trained to recognize CAPTCHA images, achieving high accuracy.
