This article presents a complete citrus detection system using YOLOv8, FastAPI, and Vue3, featuring image/video detection, ablation experiments, negative sample training, and full deployment instructions for agricultural AI applications.
This article introduces YOLO26, a new NMS‑free, lightweight computer‑vision model from Ultralytics, details its architecture, training innovations, performance benchmarks, multi‑task capabilities, and provides a step‑by‑step Python tutorial for installing the library, running a Hello‑World inference, and interpreting the detailed latency breakdown.
The paper introduces LEVIRDet‑159, the largest unified remote‑sensing detection dataset with 159 categories and 2.5 M annotations, and its foundation model LEVIRDetNet, which achieves state‑of‑the‑art performance on nine external benchmarks after a single training run, demonstrating strong cross‑scene generalization.
By fine‑tuning Amazon Nova Lite 1.0 on Amazon Bedrock, the study demonstrates how a small training dataset can dramatically improve instruction following and reduce detection boxes—up to 92% fewer—while achieving Pro‑grade accuracy in aerial group detection and low‑light monitoring, all at a fraction of the cost.
Beyond-Hungarian replaces the costly Hungarian assignment in DETR with a differentiable, query‑free matching scheme that halves training latency, boosts large‑object AP by 4.2 points, and introduces a GT‑Probe module and dual‑loss framework, while detailing trade‑offs, ablations, and future challenges.
The article dissects the query‑activation imbalance in DETR‑based detectors, explains PaQ‑DETR’s pattern‑sharing and quality‑aware assignment mechanisms, and shows how these jointly raise detection mAP by up to 4.2% on COCO with less than 5% extra FLOPs.
This tutorial walks readers through the fundamentals of YOLOv12, covering model variants, dataset preparation with Roboflow, optional FlashAttention acceleration, installation, model selection, training commands, post‑training tasks such as tracking, validation, inference, exporting to ONNX, and benchmarking, all with concrete code snippets and practical tips.
This article provides a comprehensive overview of the YOLO (You Only Look Once) algorithm, explaining its core principles, architecture, version history, training workflow, real‑world applications, strengths, and current limitations for modern computer‑vision tasks.
The article explains how YOLOv3’s architectural upgrades—Darknet‑53 backbone, three‑scale feature fusion, refined anchors and multi‑label classification, plus dynamic input sizing—enable a pig‑recognition model trained on 2,456 images to achieve up to 20% higher detection rates and AP scores of 0.673–0.981.
YOLO26, the latest Ultralytics release, introduces a unified model family with five sizes, removes distribution focal loss, offers end‑to‑end inference without NMS, adds progressive loss balancing and the MuSGD optimizer, and delivers up to 43% faster CPU performance, making it ideal for edge and real‑world vision applications.
This article walks through YOLOv2’s eight core upgrades over YOLOv1, explains the design rationale behind each change, provides detailed PyTorch code for the backbone, neck, head and prediction layers, demonstrates training on COCO, and outlines further optimization directions for real‑world object detection.
This guide walks through installing OpenCV and php-facedetect, writing PHP code to capture webcam images, crop them, run Pico face detection, and overlay bounding boxes, providing a complete example for object detection using PHP.
This article provides a comprehensive, hands‑on walkthrough of YOLOv1—from its single‑stage detection principles and core architectural questions to a full PyTorch implementation, training pipeline, common pitfalls, and a live camera demo—targeted at developers transitioning into AI.
This article introduces eight noteworthy open‑source projects—including an interactive prompt‑engineering tutorial, Claude Cookbooks, an offline speech‑to‑text tool, an eBook‑to‑audiobook converter, the Servo browser engine, a free programming‑books collection, a real‑time object‑detection model, and other popular repositories—each with brief descriptions and GitHub links.
This article presents a curated list of eight widely used remote sensing object detection datasets covering indoor scenes, landslides, drone imagery, crop diseases, safety vests, human fractures, urban issues, and plant diseases, each with size estimates and direct download links for researchers.
The article outlines four key innovation avenues—architecture redesign, task expansion, information fusion, and paradigm shift—highlighting recent works such as Mr. DETR, Grounding DINO 1.5, SM3Det, and RoboFusion, and offers a curated list of 176 cutting‑edge object‑detection papers with code and datasets for free.
This article explains PHP’s is_object() function, detailing its syntax, parameters, and return values, and demonstrates through code examples how to check whether variables such as objects and arrays are objects, helping developers avoid type errors at runtime.
This article systematically reviews every YOLO version, classifies five major improvement directions—architecture enhancements, efficiency optimizations, multi‑task learning, temporal modeling, and domain‑specific customizations—provides concrete paper references, code links, and dataset resources to help researchers and engineers quickly locate and apply the most effective techniques.
The UPRE framework introduces multi‑view domain prompts and unified representation enhancement to achieve zero‑shot domain adaptation for object detection, dramatically improving detection accuracy on unseen target domains across diverse visual scenarios.
The UPRE paper, presented at ICCV, introduces a multi‑view domain prompt and a unified representation enhancement to enable zero‑shot domain adaptation for object detection, achieving state‑of‑the‑art performance across diverse weather, geographic, and synthetic‑to‑real scenarios.
This tutorial walks through Python-based computer‑vision techniques—including XOR‑based image encryption, mask and ROI methods, digital watermark embedding via bit‑plane and LSB, sensitivity‑driven object detection, and HSV‑based green‑screen removal—providing complete code snippets and practical guidance for rapid AI‑assisted learning.
VisionReasoner presents a reinforcement‑learning‑driven unified framework that simultaneously tackles detection, segmentation, and counting tasks, employing a novel multi‑target cognition strategy and efficient Hungarian‑based matching, and demonstrates substantial gains—29.1% on COCO detection, 22.1% on ReasonSeg, and 15.3% on CountBench—using only 7,000 training samples.
VisionReasoner introduces a reinforcement‑learning‑driven unified framework that simultaneously handles detection, segmentation, and counting tasks within a single model, achieving 29.1% higher COCO detection AP, 22.1% better ReasonSeg segmentation, and 15.3% improvement on CountBench, while requiring only 7,000 training samples and offering efficient multi‑target matching via batch computation and the Hungarian algorithm.
YOLOE unifies object detection and segmentation in a single efficient model that supports text, visual, and prompt‑free inference, introduces RepRTA, SAVPE, and LRPC strategies, and achieves higher AP with up to three‑fold lower training cost and 1.4× faster inference on GPUs and mobile devices, as demonstrated by extensive LVIS and COCO experiments.
This paper presents HGO‑YOLO, a lightweight real‑time anomaly‑behavior detector that integrates HGNetv2 and GhostConv into YOLOv8, achieving 87.4% mAP with just 4.6 MB of parameters and 56 FPS on CPU, and validates its performance across multiple datasets and hardware platforms.
DefMamba introduces a multi‑scale backbone, deformable Mamba modules, and a dynamic scanning strategy to preserve image spatial structure, achieving state‑of‑the‑art performance on image classification, object detection, and semantic segmentation benchmarks.
The article demonstrates how ByteDance’s new Doubao‑1.5 multimodal model can be used to locate and count objects in images—such as sushi plates, street signs, and cartoon hats—by generating coordinates and overlaying visual annotations through a concise Python script.
The paper introduces YOLOE, a new YOLO‑based model that supports text, visual, and no‑prompt open‑world detection and segmentation, detailing its lightweight RepRTA, SAVPE, and LRPC modules and showing benchmark gains in speed and zero‑shot performance on LVIS and COCO.
The NK‑Remote repository introduces LSKNet and Strip R‑CNN, two lightweight yet powerful models for remote‑sensing object detection that dynamically adjust receptive fields and combine square‑and‑strip convolutions, achieving state‑of‑the‑art performance on benchmarks such as DOTA, FAIR1M, HRSC2016, and DIOR.
This paper proposes a multi‑teacher distillation framework for continual learning that combines active data rehearsal with feature‑decoupled distillation, demonstrating superior performance on PASCAL VOC and COCO benchmarks while mitigating catastrophic forgetting and balancing stability‑plasticity trade‑offs.
D-FINE introduces Fine-grained Distribution Refinement and Global Optimal Localization Self-Distillation to overhaul DETR's bounding-box regression, achieving 54‑59% AP on COCO and Objects365 at 78‑124 FPS while surpassing YOLO and RT-DETR in both accuracy and speed.
YOLOv12 introduces an attention‑centric architecture, a lightweight regional attention module, and the R‑ELAN aggregation network, delivering consistent mAP gains and lower latency across N, S, M, L and X model scales while surpassing previous YOLO versions and other real‑time detectors.
The article introduces Microsoft’s open‑source Computer Vision Recipes library, describing its purpose, target audience, repository links, supported vision scenarios such as image classification, similarity, detection, key‑point, segmentation, action recognition, multi‑object tracking and crowd counting, and provides guidance on using PyTorch, Azure and GPU resources.
This article explains how to install OpenCV and php-facedetect libraries, write PHP code to capture images from a webcam, perform face detection using the pico library, and display the results, providing a step‑by‑step guide for object detection with PHP.
This comprehensive tutorial walks you through YOLOv8 object detection—from environment setup and dataset preparation to model training, validation, testing, and conversion to ONNX and TensorRT—providing clear commands, code snippets, and visual results for each step.
This tutorial explains how to install OpenCV and php-facedetect, write PHP code to capture images from a webcam, perform face detection, and display the results, providing step‑by‑step commands and a complete example script.
This article provides an in‑depth overview of the Faster R‑CNN object detection framework, covering its background, key innovations such as the Region Proposal Network, detailed algorithmic principles, training procedures, experimental results on PASCAL VOC and MS COCO, and a reproducible PyTorch implementation.
Leveraging the YOLO family of deep‑learning models, this study demonstrates efficient filtering of irrelevant medical images, accurate classification of textual reports, and robust detection of occluding objects, achieving high precision and speed on both CPU and GPU, while outlining training details, performance metrics, and future improvements.
This guide demonstrates how to build a web‑based real‑time dog detector that accesses the phone camera via the browser, processes video frames with TensorFlow.js and a pre‑trained COCO‑SSD MobileNet V2 model, and plays an audio alert when a dog is recognized, all deployed on an Android device using Termux.
This tutorial explains how to install required libraries, write PHP code that captures images from a webcam, uses OpenCV and php‑facedetect to detect faces, and displays the results with annotated bounding boxes, providing a foundation for further object detection projects.
This article introduces M2Doc, a plug‑in multimodal fusion approach that equips visual‑only object detectors with textual and semantic awareness, detailing its early‑ and late‑fusion modules, experimental validation on DocLayNet, M6Doc and PubLayNet, and future research directions.
This article reviews traditional computer‑vision tasks—classification, detection, and segmentation—highlights their limitations, introduces open‑vocabulary detection and segment‑anything models such as GLIP, Grounding DINO, and SAM, and details how Huolala applies these advances to driver‑license, packing, and vehicle‑sticker inspections for safer, more efficient AI‑driven operations.
The article reviews the Open Vocabulary Detection Contest organized by the Chinese Society of Image and Graphics and 360 AI Institute, describing the competition setup, dataset characteristics, and detailed winning approaches that combine Detic, CLIP, prompt learning, and multi‑stage pipelines to achieve strong few‑shot and zero‑shot object detection performance.
The Scale‑Aware Modulation Transformer (SMT) introduces a lightweight SAM module and an Evolutionary Hybrid Network that together achieve higher accuracy on ImageNet, COCO, and ADE20K while using significantly fewer parameters and FLOPs than existing CNN and Transformer baselines.
This article explores Huolala's deployment of mobile AI image algorithms for driver document verification and vehicle sticker inspection, detailing model design, lightweighting, hybrid processing, data stream handling, and on‑device deployment that boost efficiency, privacy, and real‑time performance in logistics operations.
This article introduces the YOLOv8 object‑detection framework on the Ultralytics platform, covering environment setup, command‑line and Python APIs for inference, model‑file options, result interpretation, data annotation, training procedures, and exporting models to various deployment formats.
At CVPR 2023 the Xiaohongshu team presented OvarNet, a unified one‑stage Faster‑RCNN model built on CLIP that uses prompt learning and knowledge distillation to jointly detect objects and recognize open‑vocabulary attributes, achieving state‑of‑the‑art results on VAW, MS‑COCO, LSA and OVAD datasets.
This article reviews the original DETR model, analyzes its strengths and weaknesses, and then examines two major follow‑up works—Deformable‑DETR and DAB‑DETR—explaining how they modify attention mechanisms, introduce deformable convolutions and dynamic anchor boxes to accelerate convergence and improve small‑object detection.
This article reviews the evolution of open‑vocabulary object detection, describing the OVR‑CNN paradigm, the RegionCLIP enhancements, and the CORA model with region prompting and anchor pre‑matching, and discusses their impact on future multimodal AI systems.
This article introduces DAMO‑YOLO, a high‑performance object detection framework that combines low‑cost model customization via MAE‑NAS, an Efficient RepGFPN with HeavyNeck for superior multi‑scale detection, and a full‑scale distillation technique, delivering faster inference, lower FLOPs, and higher accuracy across diverse industrial scenarios.
This article presents a comprehensive overview of ModelScope's computer‑vision models, detailing visual detection and keypoint solutions—including VitDet, YOLOX, res2net, HRNet, and 3D pose models—their architectures, performance highlights, real‑world applications, and future development plans.
This article walks through the complete workflow of building an iOS object‑detection model with Apple’s CreateML, covering data collection, JSON annotation, using Roboflow for labeling, configuring training parameters, exporting the model, and integrating it into a Swift app via the Vision framework.
ARSL, an ambiguity‑resistant semi‑supervised learning framework for single‑stage object detection, introduces Joint‑Confidence Estimation and Task‑Separation Assignment to resolve selection and assignment ambiguities in pseudo‑labels, thereby markedly improving pseudo‑label quality and achieving state‑of‑the‑art AP gains on COCO benchmarks.
This article walks through the complete workflow of building an iOS object‑detection model with Apple CreateML, covering data collection, JSON‑based annotation (using Roboflow), training configuration, evaluation metrics, model export, and integration into a Swift app via the Vision framework.
PaddleDetection v2.6 expands the PP‑YOLOE family with rotating, small‑object, dense‑object, and ultra‑lightweight edge‑GPU models, upgrades PP‑Human and PP‑Vehicle toolboxes, releases semi‑supervised, few‑shot and distillation learning methods, adds numerous state‑of‑the‑art algorithms, and improves infrastructure with Python 3.10, EMA filtering and AdamW support.
This article provides a comprehensive overview of object detection, describing traditional sliding‑window approaches, deep‑learning based two‑stage and one‑stage models such as R‑CNN, Faster R‑CNN, YOLO series, and discusses current challenges, improvement directions, and future research trends in the field.
YOLOv6 3.0 raises real‑time object detection performance to a new peak with 57.2% AP and 29 FPS on a T4 GPU, surpassing YOLOv7‑E6E, and introduces RepBi‑PAN Neck, Anchor‑Aided Training, and Decoupled Location Distillation to boost accuracy and efficiency.
This article explores how artificial intelligence techniques such as deep learning, object detection, multi‑object tracking, and coordinate projection are applied to football video analysis to automatically detect the ball and players, map their positions onto the field, and recognize key events like shots and goals.
This article introduces Apple’s CreateML tool, explains its supported model types, shows how to prepare and augment data, provides a Node.js script for generating synthetic training sets, and walks through training, testing, and integrating an object‑detection model into an iOS app.
This article introduces EasyCV, an open‑source all‑in‑one visual algorithm platform that abstracts diverse data sources, provides SOTA self‑supervised models, and offers ready‑to‑download datasets for image classification, object detection, segmentation, and pose estimation, complete with configuration examples.
This article describes a comprehensive solution for detecting checkboxes in document images and determining their selected or unselected status by combining YOLOv5 object detection, synthetic and semi‑synthetic data generation, specialized post‑processing, and association logic to handle varied shapes, positions, and markings.
The new YOLOv6 2.0 release upgrades lightweight and medium‑large models with a CSPStackRep backbone, self‑distillation, and a custom quantization pipeline, delivering up to 869 FPS for the quantized YOLOv6‑S and achieving 49.5%/52.5% AP on COCO while halving training time.
This guide explains how to combine the open‑source labeling platform Label Studio with the YOLOv5 object‑detection model to achieve semi‑automatic annotation, covering installation of both tools, project creation, dataset configuration, and training a custom YOLOv5 model on your own data.
YOLOv6, developed by Meituan's Vision Intelligence team, introduces a hardware‑friendly backbone, an efficient decoupled head, and advanced training strategies that together achieve up to 35.0% AP at 1242 FPS on COCO while outperforming YOLOv5, YOLOX and other same‑size models across multiple deployment platforms.
This guide explains how to replace YOLOv4's CSPdarknet53 backbone with MobileNetV1, V2, or V3 networks, detailing the architecture analysis, code implementations, training setup, dataset preparation, and inference procedures for building a lightweight object detection model.
The article reviews the Context R-CNN paper, which introduces short‑term and long‑term memory banks and an attention mechanism to incorporate temporal context from multiple frames captured by a fixed camera, enabling robust detection of partially occluded, low‑light, distant, or background‑cluttered objects, and shows quantitative gains over standard Faster R‑CNN.
This article provides a step‑by‑step guide to building, training, and evaluating a custom object detection model using PyTorch Faster R‑CNN on a microcontroller dataset, covering data preparation, configuration, model modification, training loops, loss visualization, and inference on new images.
This article provides a comprehensive guide on YOLOv5, covering its background from YOLOv3, detailed code analysis of the model architecture, step‑by‑step instructions for running detect.py, configuring yolov5s.yaml, exporting the model to TorchScript JIT and ONNX formats, and practical inference examples using PyTorch and ONNX Runtime.
This tutorial demonstrates how to perform modern object detection using the ImageAI library with only ten lines of Python code, covering the underlying computer‑vision concepts, required dependencies, step‑by‑step installation, and detailed explanation of each code segment.
At ACM MM 2021, Alibaba’s Moku Lab presented four cutting‑edge studies: an interactive video inpainting system using user doodles, a decoupled IoU regression model for object detection, a spatio‑temporal distortion‑aware video quality assessment framework, and a multimodal emotional relationship recognition dataset and benchmark.
This article details the evolution of intelligent vending cabinets, comparing RFID, gravity, dynamic and static vision solutions, and explains how deep‑learning models, data pipelines, and system architectures enable high‑accuracy, low‑loss product detection and automated operations in modern unmanned retail.
This article explains the challenges of scene text detection and introduces two state‑of‑the‑art models, ContourNet and CenterNet, detailing their architectural innovations, loss functions, and how they overcome issues like extreme aspect ratios and anchor‑based inefficiencies.
This article details Alibaba's AliMama brand risk identification system, covering the challenges of counterfeit detection, the construction of large‑scale brand datasets, the design of classification, logo detection, and variation models, their optimization, evaluation metrics, and future directions for AI‑driven brand protection.
This article presents a method that replaces traditional manual video frame quality checks with an automated YOLOv5‑based object detection pipeline, detailing data labeling, model training, loss computation, inference code, and experimental results that show higher accuracy than ResNet for classifying black, color‑screen, and normal frames.
This article presents a method that replaces manual visual inspection with an automated YOLOv5‑based object detection pipeline to classify video frames as normal, colorful, or black screens, detailing data annotation, training, loss calculation, inference code, and showing a 97% accuracy improvement over ResNet.
This article describes how 58.com built an end‑to‑end deep‑learning watermark detection service, covering business needs, data collection and augmentation, model selection and iterative improvements (Faster‑RCNN, SSD, YOLOv3, anchor‑free methods), deployment results, and future research directions.
This article introduces the fundamentals of object detection, explains the latest YOLOv4 and YOLOv5 architectures, and details the essential components—including data preparation, regularization, backbone, neck, and prediction innovations—along with label smoothing and advanced loss functions for improved detection performance.
This lecture details the evolution of product recognition algorithms from traditional image classification to deep‑learning‑based object detection, discusses challenges in dense retail scenes, presents solutions like rotated bounding boxes and multi‑source sensor fusion, and explains practical deployment in digital and unmanned stores.
This article explains how AI‑driven precise customer‑flow algorithms—covering pedestrian detection, full‑scene tracking, and person re‑identification—enable accurate offline traffic analysis, real‑time shopper profiling, and data‑driven store management for modern retail environments.
To enhance object detection robustness, the article discusses various data augmentation methods—including rotation, flipping, random cropping, scaling, color jitter, blurring, transparency adjustment, and image partitioning—providing code examples and illustrating their impact on model performance with before‑and‑after results.
This tutorial walks you through using Pipcook to train an object‑detection model that automatically identifies and locates front‑end UI components in screenshots, covering data preparation in Pascal VOC format, pipeline configuration, model training, and inference with sample code.
This article explains how to use deep‑learning object detection to automatically recognize UI components in design drafts, generate a smart JSON description, and convert it into component‑based front‑end code, covering problem analysis, dataset preparation, algorithm selection, model training, evaluation, and deployment.
CenterMask is a single‑shot, anchor‑free instance segmentation framework that predicts a coarse shape from each object’s center point and a full‑image saliency map, multiplies them to produce precise masks, and achieves competitive COCO AP while running faster than two‑stage methods like Mask R-CNN.
This paper presents a deep learning-based approach for selecting cover images in Autohome forum articles, employing Faster R-CNN for object detection, Mask R-CNN for human keypoint detection, and MobileNetV2 for attribute recognition, achieving an overall accuracy of 81.5%.
The paper presents a lightweight, anchor‑free CenterNet‑based object‑ness detector for WeChat’s Scan feature, built on a ShuffleNetV2 backbone with enlarged 5×5 depth‑wise convolutions, a streamlined detection head, and a Pyramid Interpolation Module, then quantized, ONNX‑converted and NCNN‑deployed to achieve a 436 KB model running in ~15 ms per frame on an iPhone 8 CPU.
This article reviews perception technologies used in autonomous heavy‑truck systems—including lane‑line detection, obstacle detection, and LiDAR sensing—detailing traditional and deep‑learning approaches, practical challenges on high‑speed highways, and the cost, performance, and reliability issues faced when moving these solutions to mass production.
This article details how the 58 Used‑Car mobile team built, trained, and optimized TensorFlow‑based object‑detection models for on‑device vehicle and VIN code recognition, covering data preparation, model conversion to TF‑Lite, performance improvements, engineering integration on Android/iOS, and real‑world deployment results.
WeChat’s iOS Scan‑to‑Identify feature lets users point a camera at any product or scene to instantly retrieve related e‑commerce, encyclopedia or news content, using a four‑pipeline architecture that builds massive annotated and deduplicated databases, advanced RetinaNet‑based detection, multi‑task metric learning, and scalable training, deployment and scheduling platforms, with plans to extend into domains like facial, vehicle and plant recognition.
This article details how Alibaba leveraged AI-driven hand‑gesture detection and a lightweight SSD‑based object detection model to create an interactive rock‑paper‑scissors game for Double‑11, addressing challenges of undefined gestures, real‑time mobile performance, and data collection, and achieving over 16 million page views and high accuracy.
This article introduces the fundamentals of object detection, explains key models such as R-CNN, Fast R-CNN, Faster R-CNN, YOLO, and SSD, and showcases how Alibaba's AI technology is applied to photovoltaic quality inspection to boost efficiency and accuracy in industry.
This article presents a comprehensive front‑fusion recognition pipeline for high‑definition map static obstacle detection, detailing depth‑aware mapping, precise multi‑sensor calibration, point‑cloud registration, and semi‑supervised learning techniques that improve detection accuracy over traditional image‑only methods.
The talk analyzes sample imbalance and importance in object detection, proposes IoU‑balanced negative sampling and instance‑balanced positive sampling, introduces the Prime Sample concept with Hierarchical Local Rank, and presents Importance‑based Sample Reweighting and Classification‑Aware Regression Loss, achieving consistent mAP gains without extra overhead.
AMap’s ground‑marking recognition has progressed from simple threshold methods to advanced deep‑learning pipelines—including two‑stage R‑FCN, cascade detectors with local regression, corner‑point and segmentation hybrids, and LiDAR‑based 3‑D PointRCNN—achieving over 99 % recall and sub‑5 cm positional accuracy for high‑precision map production.
This study presents a deep‑learning pipeline that employs a Cascade‑RCNN with Feature Pyramid Network to automatically detect and refine modules and their internal elements on Taobao’s 99‑promotion pages, achieving roughly 98 % precision and recall on a thousand‑image validation set and paving the way for broader e‑commerce event applications.
This guide walks you through configuring Huawei Cloud ModelArts, creating an AutoML image‑classification project, labeling cat and leopard photos, training a model, deploying it as an online service, and testing predictions to accurately differentiate between leopards and jaguars.
The proposed UI2CODE pipeline first recalls UI elements with an object detector, then uses gradient cues to separate simple from complex regions and applies an SRGAN to restore foreground details in challenging backgrounds, achieving higher precision, recall, and localization than GrabCut and Deeplab, though it demands extensive multi‑scale training data.
This article outlines Pony.ai's perception pipeline, comparing traditional and deep‑learning approaches for 2D and 3D object detection, detailing sensor fusion, detection methods, challenges such as occlusion and distance estimation, and how hybrid techniques improve accuracy for autonomous driving.
By integrating YOLOv3 video object detection into Youku’s ad platform, the team replaced traditional subtitle‑based and scene‑based placements with precise object‑level targeting, achieving higher relevance, expanded inventory, and a 20% click‑through increase despite 3.5× higher exposure.
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.
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.