This article summarizes a team's practical experience with various image feature extraction methods—including global, local, and CNN features—and demonstrates their use in image retrieval and no‑reference quality assessment through extensive experiments and analysis.
Ren Haibing’s Youku Cognitive Lab talk reviews object segmentation’s motivation, explains semantic and instance concepts, presents UNet‑based and category‑agnostic methods—including fast video segmentation with motion cues—and reports high IoU results while outlining future edge‑aware, label‑free, and non‑online video segmentation research directions.
The article describes an intelligent cover‑image selection system for NetEase News that automatically filters unsuitable illustrations, assesses image quality with a pairwise‑trained deep model across clarity, color and composition, and smartly crops images using aspect‑ratio‑aware object detection, dramatically cutting manual editing and enabling confidence‑based automatic publishing.
The article outlines Tencent Cloud’s face‑recognition technology—from its deep‑learning‑based algorithm training and multi‑layer system architecture, through the YouTu Lab‑powered product suite for detection, analysis, comparison, liveness and search, to real‑world deployments in security, metro transportation and retail, highlighting integration challenges and performance optimizations.
This article provides a concise introduction to image processing, covering its hierarchical structure, fundamental techniques such as classification, detection, segmentation, geometric transformation, and the most widely used libraries and deep‑learning frameworks for newcomers.
This article summarizes a computer‑vision salon where Dr. Ji Yongnan explains imaging pipelines, traditional feature‑based methods, deep‑learning breakthroughs, Tencent Cloud AI services, real‑world case studies, and answers audience questions about machine‑vision versus computer‑vision and data‑scarcity challenges.
The YOUKU‑VSRE 2019 challenge invites researchers to develop state‑of‑the‑art video enhancement and super‑resolution models using the largest, most diverse simulated‑noise dataset, with three competition stages (preliminary, semi‑final, final), cash prizes up to ¥100,000, certificates, and fast‑track recruitment opportunities at Alibaba (Youku).
The CVPR 2019 WAD Autonomous Driving Challenge, hosted in Long Beach, introduces four new tasks—including object‑detection and tracking transfer‑learning tracks using Didi’s massive D²‑City and Berkeley’s BDD100K datasets, plus a large‑scale detection interpolation track—aimed at advancing vision algorithms under diverse, difficult driving conditions, with global teams invited to register by May 31 and winners announced at the workshop on June 17.
This article introduces how machine learning problems are formulated as optimization tasks, explains the construction of objective functions with examples such as linear regression, robust fitting, regularization, and demonstrates various applications ranging from K‑means clustering to image inpainting and 3D reconstruction.
This article provides an extensive review of deep learning techniques applied to computer vision, covering the evolution of CNN architectures, image and video processing tasks, 2.5‑D and 3‑D reconstruction, object detection, segmentation, tracking, SLAM, and various practical applications such as AR, content retrieval, and autonomous driving.
This article explains why simple MD5 checks cannot assess image similarity and introduces two major approaches—image hashing and image feature extraction—detailing their algorithms, practical performance, and how to compare images efficiently using Hamming distance and indexing techniques.
This article presents a comprehensive AI‑driven framework that fuses video‑based customer action detection, RFID‑based product flip detection, and bipartite graph matching to accurately determine when, where, and which customer interacts with which SKU in a retail environment, discussing algorithms, optimizations, and experimental results.
This article describes a comprehensive, end‑to‑end AI visual algorithm platform that unifies data collection, annotation, model training, deployment, testing, quality evaluation, and service gateways, illustrating how such integration improves transparency, efficiency, and quality across use cases like background removal, face swapping, and clothing recommendation.
This article traces the evolution of convolutional neural networks from the pioneering AlexNet through VGG, Inception, ResNet, Inception‑v4, Inception‑ResNet and ResNeXt, highlighting architectural innovations, performance gains, and the underlying biological inspirations that shaped modern deep learning models.
This article explains why AI‑generated faces from StyleGAN are hard to distinguish, introduces an online game for testing realism, and provides practical visual cues—such as water spots, background errors, asymmetric glasses, hair artifacts, and teeth anomalies—to reliably identify fake images.
This article presents an overview of computer‑vision based OCR in Ctrip's operations, detailing deep‑learning text detection methods for controlled and uncontrolled scenarios, sequence‑based recognition models, training strategies with synthetic data, and performance results, while discussing current challenges and future improvements.
The UI2CODE system tackles UI layout analysis by first extracting backgrounds with Sobel, Laplacian and Canny edge detection plus a flood‑fill algorithm, then isolating foreground components through connected‑component analysis and a Faster R‑CNN classifier, and finally fusing both pipelines to achieve superior precision, recall and IoU on Xianyu app screenshots.
This article introduces core image processing concepts—pixel fundamentals, binary, grayscale, and RGB images, matrix representation—and demonstrates practical implementations of cropping, canvas creation, watermarking, translation, rotation, and scaling using Python's OpenCV and NumPy libraries, including algorithm choices for resizing.
A data scientist from Besedo has open‑sourced a massive NSFW image dataset containing 1.589 million pictures, organized into 159 primary categories and further sub‑categories, with download scripts and GitHub links, requiring about 500 GB of storage and cautioning against viewing in the office.
The article recounts how a 1960 Playboy portrait of Lena Söderberg was adopted by image‑processing researchers as a standard test image, explains the technical and cultural reasons for its lasting popularity, and follows her unexpected rise to fame within the scientific community.
Data scientist Evgeny Bazarov has open‑sourced a 1.589 million‑image NSFW dataset organized into 159 fine‑grained categories, providing GitHub links, download scripts, and a 500 GB storage requirement, enabling researchers to build more precise adult‑content detection models.
After the Chinese New Year break, this curated collection of key AI articles—spanning computer vision, speech recognition, natural language processing, recommendation systems, and more—helps technical readers quickly regain momentum in work and study by revisiting core technologies with real‑world case studies.
This tutorial walks through using a webcam, Mask R‑CNN, and Python to automatically detect available parking spaces, track stationary vehicles, compute Intersection‑over‑Union to confirm emptiness, and send SMS alerts via Twilio, providing full code snippets and practical tips.
At AAAI 2019 in Hawaii, JD AI Research Institute had eight papers accepted covering machine learning, natural language processing, computer vision, and multimodal AI, highlighting innovations such as AutoZOOM black‑box attacks, SACN for knowledge base completion, and temporally aware video captioning models.
Alibaba’s search division showcases how AI transforms traditional retail by digitizing in‑store foot traffic, employing camera‑based person detection, re‑identification, RFID‑enhanced product interaction, and edge‑optimized models to generate real‑time customer insights, heatmaps, and personalized recommendations that bridge offline and online shopping experiences.
This guide introduces the open‑source Python library face_recognition, explains its high‑accuracy (up to 99.38%) facial detection and landmark capabilities, provides step‑by‑step code examples for locating faces, extracting landmarks, and comparing identities, and lists practical use‑case scenarios and the GitHub repository.
This article provides a comprehensive overview of computer vision, covering its basic concepts, traditional image processing techniques, modern deep‑learning approaches, real‑world AI application cases, and the cloud infrastructure needed to support large‑scale deployment, while also offering skill‑advancement guidance.
This article details Alibaba's end‑to‑end solution for digitizing offline retail foot traffic, combining existing surveillance cameras, RFID tags, and advanced AI techniques such as lightweight YOLO detection, knowledge distillation, and multi‑level pedestrian re‑identification to capture, analyze, and act on shopper behavior for both business operations and personalized in‑store experiences.
This week’s Alibaba tech roundup highlights the DAMO Academy’s 2019 top‑10 technology trends—from smart cities and AI chips to blockchain and 5G—plus breakthrough AI liver‑tumor segmentation results, the open‑source Fusion design system, a Flink Forward China recap, a new computer‑vision paper collection, and an upcoming Apache Dubbo live session.
This presentation outlines the fundamentals of computer vision, discusses key factors such as algorithm research, large‑scale training platforms, intelligent data processing, and hardware optimization, and shares practical experiences from DeepGlint on building a world‑class visual AI system and its real‑world applications.
The talk provides a comprehensive overview of computer vision, defining its scope, detailing low‑, mid‑, and high‑level processing pipelines, reviewing classic filters and feature extractors, explaining deep‑learning breakthroughs such as CNNs and YOLO, and showcasing Tencent Cloud AI services, career paths, and learning resources.
This article presents a technical solution for extracting representative key frames from time‑series screenshots of a mobile app installation process, covering pixel sampling, dimensionality reduction, classic feature extractors (SIFT, HOG, ORB), auto‑encoder based deep learning, and clustering methods such as KMeans and DBSCAN, along with practical results and performance analysis.
The article surveys the nineteen leading AI technologies—from natural language generation and speech recognition to digital twins and marketing automation—detailing their core functions, common use cases such as customer service, security, content creation, and the key vendors delivering each solution.
This comprehensive overview examines the fundamentals, historical milestones, key algorithms, major datasets, policy support, industry applications, and future trends of face recognition technology, highlighting its rapid growth within computer vision and artificial intelligence.
This article presents a step‑by‑step solution that combines computer‑vision preprocessing, OCR, CNN classification, shape matching, and inpainting to isolate meaningful foreground elements from images with intricate backgrounds, discussing practical results, limitations, and code implementations.
This tutorial walks through extending a deep‑learning object detector to process live video streams using OpenCV and Python, covering setup, command‑line arguments, model loading, frame‑by‑frame detection, drawing bounding boxes, FPS measurement, and performance tips.
This tutorial explains how to capture screen images from GTA V (or similar games) using Python and OpenCV, covering screen grabbing, converting to NumPy arrays, handling performance, and setting up basic input simulation to enable AI-driven autonomous driving experiments.
Alibaba’s senior tech expert explains the concept, architecture, core AI capabilities, real‑world case studies, and future roadmap of the Tmall “Future Store”, a vision‑driven, sensor‑rich unmanned retail experience that merges computer‑vision, edge computing, and data‑driven operations.
The article introduces convolutional neural networks, explains core concepts such as convolution, padding, stride, and pooling, demonstrates how to calculate output dimensions, and provides a step‑by‑step Keras example that builds, compiles, and trains a multi‑layer CNN for image classification.
The article summarizes Jifeng Dai's 2018 AI Pioneer talk on recent object‑detection breakthroughs, detailing R‑FCN and its extensions, Deformable ConvNets, video object detection techniques, and concluding remarks on remaining challenges in large‑scale and mobile vision.
Deep neural networks, propelled by breakthroughs such as AlexNet and advances in GPU and TPU hardware, are rapidly moving from academic research into diverse applications—including earthquake prediction, medical imaging, and autonomous driving—driving massive industry investment, new semiconductor designs, and intense competition among tech giants and startups.
Fei‑Fei Li, the pioneering AI researcher and former Google Cloud AI lead, rose from humble beginnings in China to create the ImageNet dataset, drive breakthroughs in computer vision, and now returns to Stanford, illustrating how curiosity and perseverance can transform both academia and industry.
This article reviews major deep‑learning object detection approaches—including one‑stage YOLO and SSD and two‑stage RCNN, Fast RCNN, and Faster RCNN—then provides a step‑by‑step explanation of Faster RCNN’s architecture, region‑proposal network, RoI pooling, loss functions, and sample PyTorch code.
Sketch2Code leverages Microsoft’s custom vision model, OCR, and Azure services to automatically convert hand‑drawn UI mockups into functional HTML code, detailing its workflow—from image upload and element prediction to layout generation and final HTML output—plus links to the repository and demo site.
This article surveys the evolution of face recognition from early statistical methods to modern deep‑learning approaches, outlines key researchers, open‑source projects, popular APIs, core processing steps, the DeepFace architecture, datasets, and experimental results, providing a comprehensive guide for practitioners and researchers.
iQIYI released iQIYI-VID, the world’s first multimodal, multi-angle celebrity video dataset (1,000 hours, 500,000 clips, 5,000 celebrities) for a new AI competition focusing on multimodal video person recognition, which has attracted global university teams and top computer‑vision judges to advance AI understanding in entertainment.
This article reviews the evolution of visual object detection, explaining traditional region‑based approaches, the rise of deep‑learning two‑stage frameworks such as R‑CNN, Fast R‑CNN and Faster R‑CNN, and the faster one‑stage models like Overfeat, YOLO, SSD and RetinaNet, together with their design choices, training strategies and loss functions.
Tencent's AI Breast System combines mammography, pathology, MRI and ultrasound analysis using a multi‑scale, progressive TMuNet model that processes four views, learns from physician feedback, and delivers lesion localization, malignancy scoring and automated reports, achieving up to 92% sensitivity and reducing annotation time.
The July 28 Tencent Cloud community salon in Beijing gathered five AI experts who demonstrated practical AI applications—including computer‑vision face filters, OCR services, smart construction attendance, game AI, and breast‑cancer detection—showing how cloud‑based models, data pipelines, and deployment strategies turn research into usable products.
This article describes Meituan's large‑scale AI‑driven image moderation platform, detailing deep‑learning based watermark detection, celebrity face recognition, pornographic image detection, and scene classification techniques, along with system architecture, data preparation, model evaluation, and deployment considerations.
This article explains online action prediction, the challenges of early‑stage classification, and introduces a Scale Selection Network that dynamically chooses optimal temporal windows using dilated convolutions, regression and classification sub‑networks, achieving state‑of‑the‑art results on two benchmark datasets.
This article explains how Alibaba's Image & Beauty team designs and iterates a practical fashion image dataset by aligning data purpose, integrating professional knowledge, handling sample scarcity and structured noise, and defining fine‑grained evaluation metrics to enable AI models that truly understand clothing.
The paper proposes a no‑reference image sharpness metric that computes strong‑edge validity statistics—ratio of maximum directional gradient sum to squared strong‑edge count—across image blocks, classifies them into grades, and effectively handles defocus and motion blur for applications such as video thumbnail selection.
This article walks through creating a concise 200‑line Python script that automatically detects facial landmarks with dlib, aligns faces using Procrustes analysis, corrects color differences, and blends a second face onto a first image, complete with code snippets and step‑by‑step explanations.
This tutorial explains how to implement high‑accuracy face recognition using OpenCV, Python, and deep learning by leveraging dlib's deep metric learning, creating a custom dataset, encoding facial embeddings, and performing real‑time identification on images and video streams.
Meituan’s OCR system replaces the classic preprocess‑segment‑recognize pipeline with deep‑learning components—CNN‑based text detection, synthetic‑data‑trained character models, and BLSTM‑CTC sequence recognition—delivering far higher accuracy on noisy, varied real‑world images such as menus, receipts, and IDs, though further integration with layout analysis remains needed.
The paper introduces a context‑contrast local feature and a gated multi‑scale fusion mechanism that together enhance pixel‑level scene segmentation, especially for inconspicuous objects, and validates the approach with state‑of‑the‑art results on Pascal Context, SUN‑RGBD, and COCO Stuff datasets.
The article analyzes the challenges of declining online traffic and low in‑store conversion, then details how Tencent Cloud’s AI‑driven retail suite—covering scientific site selection, commercial‑area insights, computer‑vision store management, one‑code product tracking, and brand sentiment analysis—offers data‑powered solutions for modern retailers.
This article details how Alipay's scanning technology team improved QR code recognition by refining aspect‑ratio tolerance, introducing new pattern detection modes, applying diagonal filtering, leveraging logistic‑regression classification, adjusting jump‑line intervals, and moving binarization to GPU, resulting in a 6.95‑point increase in recognition rate and significantly reduced processing time.
This article walks through creating a real‑time card‑based augmented reality prototype using Python, OpenCV, and NumPy, covering surface detection, feature extraction, matching, homography estimation, and RANSAC to project 3D models onto a reference plane.
This article explains the fundamentals of optical flow, traces its development from early variational methods to modern deep‑learning models like FlowNet, and discusses practical applications such as video object detection, semantic segmentation, and novel view synthesis, highlighting both technical challenges and future research directions.
This article introduces how image‑recognition techniques, particularly using the Sikuli tool, can be applied to UI automation testing for both web and mobile applications, covering practical scenarios, core principles, a suite of useful functions, example code, and the advantages and limitations of the approach.
This article introduces the use of image‑recognition techniques, particularly the Sikuli tool, for UI automation testing, covering typical scenarios, underlying principles, key functions such as Find, click, wait, and type, as well as example code, and discusses the advantages and limitations of this approach.
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 article explains the fundamentals of SLAM technology, compares sparse, dense, and vision‑based approaches, evaluates four robot‑vacuum navigation schemes, presents test results across different scenarios, and discusses market opportunities for 360's laser‑SLAM driven vacuum cleaners.
This tutorial walks you through setting up a simple face‑recognition pipeline using Dlib and scikit‑image in Python, covering required tools, data preparation, the recognition workflow, and a complete 40‑line script with sample images and execution results.
This article explains how to write a compact Python script (about 200 lines) that automatically detects facial landmarks with dlib, aligns two faces using Procrustes analysis, corrects color differences, and blends the second face onto the first using OpenCV, complete with full source code and visual examples.
The article details Suning’s Beidou system, an AI‑driven solution that combines video, WiFi and facial recognition to accurately count and analyze customer flow, improve store operations, and enable intelligent services such as personalized recommendations, automated payment, and safety monitoring for modern retail environments.
The JD Big Data R&D team announced that three of its research papers—covering cross‑domain human parsing, multi‑view outlier detection, and orthogonal weight normalization for deep neural networks—were accepted at the prestigious AAAI‑2018 conference, highlighting the department's contributions to computer vision, data mining, and deep learning.
This tutorial shows how to generate a labeled dataset from the open‑source WordPress "Really Simple CAPTCHA" plugin, train a lightweight convolutional neural network using Python, OpenCV, Keras and TensorFlow, and decode real captchas within fifteen minutes, demonstrating the power of modern computer‑vision techniques.
This article explains the fundamentals of image recognition technology, its underlying principles, processing steps, neural‑network and nonlinear‑dimensionality‑reduction approaches, and highlights its wide‑range applications and future potential across many industries.
On the cusp of the 2017 year‑end, the author humorously marks the last 90‑post turning 18 while offering a Python‑OpenCV tutorial that, in just 30 lines of code, applies a beauty filter to make anyone look younger, complete with installation steps and sample output.
This article introduces how to leverage the opencv4nodejs Node.js module—binding OpenCV’s full API—to develop a face detection and recognition application, highlighting the CPU‑intensive nature of computer‑vision tasks, the limitations of JavaScript, and the availability of synchronous and asynchronous examples.
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 tutorial walks through setting up Ubuntu 17.10 with Python 2.7, installing required packages, compiling dlib, and using the face_recognition library to detect, identify, and beautify faces through multiple code examples.
This guide walks through setting up Ubuntu 17.10 with Python 2.7, installing dlib and the face_recognition library, and demonstrates five practical examples ranging from a one‑line face‑recognition command to facial feature extraction and beautification, complete with code snippets and screenshots.
This article reviews an ICCV paper introducing a hierarchical multimodal LSTM that jointly embeds images, phrases, and whole sentences, enabling detailed image descriptions and superior performance on Flickr30K, MS‑COCO, and region‑phrase datasets compared to previous methods.
Pornhub unveiled an AI model that uses computer‑vision techniques to automatically recognize and tag over ten thousand adult performers, allowing users to search more precisely while also involving human reviewers to verify and improve the system’s accuracy.
This guide introduces the easy-to-use Python face_recognition library, explains its dlib‑based deep learning accuracy, details installation on various platforms, and demonstrates command‑line and Python API usage for detecting faces, facial landmarks, and real‑time recognition.
This article surveys recent progress in deep learning by presenting key computer‑vision architectures such as AlexNet, VGG, GoogleNet, ResNet, ResNeXt, RCNN, YOLO, SqueezeNet, SegNet and GANs, providing brief descriptions, their advantages, and links to original papers and Keras implementations.
The article discusses the evolution of image search engines, outlines key challenges such as image quality, watermarks, speed, and feature extraction, and explains how the Facenet deep‑learning model with Triplet loss can be used to generate compact image embeddings for efficient similarity search.
This article explains the four main stages of the SIFT algorithm—scale‑space construction, DoG extrema detection, keypoint orientation assignment, and descriptor generation—while providing a full Java implementation with detailed code snippets and explanations of each processing step.
This article provides a detailed overview of autonomous driving, covering its evolution from electric and shared vehicles, major industry players, technical definitions, SAE level classifications, core modules such as perception, localization, decision and control, key datasets like KITTI, and emerging business opportunities in the sector.
In an exclusive interview, Alibaba AI Lab's distinguished scientist Dr. Wang Gang discusses the lab's research on multimodal AI, edge computing, AI hardware, bio‑inspired cognition, quantum‑deep‑learning integration, and the challenges of moving from recognition to true understanding, while also outlining Alibaba's AI talent recruitment plans.
This tutorial walks through using Python and OpenCV to capture screen images from GTA V, covering screen grabbing, converting to NumPy arrays, improving frame rates, handling key and controller input, and setting up a basic autonomous driving test environment, all without relying on pre‑existing libraries.
Alibaba’s AI breakthroughs—from a foot‑scanning shopping demo that lets a Google engineer instantly find matching shoes, to a record‑setting vehicle detection model on KITTI and world‑leading OCR for real‑time image review—showcase the power and commercial potential of modern computer‑vision research.
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.
The article reviews the three historic AI booms, explains why today’s cloud‑based visual intelligence represents a distinct era, outlines five key factors for successful visual AI, and showcases real‑world Alibaba Cloud applications such as product search, city‑wide monitoring, medical diagnosis, and visual advertising.
Image search, once limited by early CBIR techniques, has surged back thanks to deep learning, offering improved relevance, coverage, scalability, and user experience across applications like e‑commerce, shopping, entertainment, and surveillance, while integrating data, users, models, and systems to bridge the semantic gap.
Last month Ctrip Technology Center hosted a deep‑learning meetup featuring academic and industry experts from UCL, Fudan, Southeast University, Nanjing University, Huawei, Sogou and others, who presented real‑world applications of deep learning in travel, advertising, natural language processing, computer vision, and knowledge graphs.
This article provides a comprehensive overview of OCR technology, explains its definition and application scenarios, and details a five‑step workflow—including target extraction, preprocessing, character localization, digit matching, and format validation—specifically illustrated with bank card number recognition.
This article explains augmented reality (AR), compares it with virtual reality (VR), describes the technical pipeline and design challenges of AR, illustrates its unique interaction methods, and discusses current limitations and future possibilities, using Pokemon Go as a concrete example.
This article reviews how Ctrip’s travel guide community leverages deep learning models such as CNN, LSTM, and RCNN for multilingual text analysis, image classification, video moderation, and data matching, and outlines future directions like knowledge graphs and virtual reality.
Mobile visual search, a long‑standing dream, has evolved from early research to a production‑grade system at Pailitao, where a five‑module AI pipeline—category prediction, object detection, feature extraction, indexing, and ranking—enables billions of images to be searched instantly on mobile devices.
This tutorial shows how to use Python, OpenCV, and basic image‑processing techniques to automatically detect whether a kitchen oven is on by analyzing the red indicator light captured by a home camera, providing a simple safety alert system.
This article traces the evolution of computer vision from its 1960s origins, explains the challenges of image classification and semantic segmentation, and introduces pixel-level decision forest algorithms as an efficient solution for large‑scale pixel classification tasks.
This tutorial walks through using PHP with OpenCV to detect faces, noses, and mouths, filter out false detections, and programmatically overlay a mustache image onto the identified region, complete with step‑by‑step algorithms, debugging tips, and visual results.
This article describes a project that enhances an open‑source card‑number scanning solution to recognize 19‑digit Chinese debit cards, addressing challenges such as black‑printed fonts, light‑colored embossed fonts, background filtering, single‑character OCR, and Luhn‑based checksum verification.
