计算机视觉四大任务模型汇总

计算机视觉中有四大核心任务:

1-分类任务、2-目标检测任务、3-目标分割任务 和 4-关键点检测任务

文章1:

一文读懂计算机视觉4大任务

文章2:

图像的目标分割任务:语义分割和实例分割

不同任务之间相关但不完全相同,因此不同的任务最好选择相应的模型,话不多说,看表:

注:表中github链接并不一定是模型的正式版本,只是本文用于展示模型的网络结构和应用

1-分类任务模型

|--------|------------------|-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| 序号 | 模型 | ipynb模型的github链接 |
| 1 | LeNet | https://github.com/udacity/CarND-LeNet-Lab |
| 2 | AlexNet | https://github.com/Fannjh/AlexNet-TF |
| 3 | VGGNet | https://github.com/Fozan-Talat/Image-Classifier-VGG |
| 4 | GoogLeNet | GitHub - AbdelrahmanShehata482/CNN-project: CNN_Project (py and ipynb code ) (Vgg16-GoogleNet from scratch) |
| 5 | ResNet | GitHub - ry/tensorflow-resnet: ResNet model in TensorFlow |
| 6 | DenseNet | GitHub - titu1994/DenseNet: DenseNet implementation in Keras |
| 7 | MobileNet | https://github.com/Zehaos/MobileNet |
| 8 | EfficientNet | https://github.com/qubvel/efficientnet |
| 9 | SVM(支持向量机) | https://github.com/Think103/- |

2-目标检测任务模型

|--------|---------------------|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| 序号 | 模型 | ipynb模型的github链接 |
| 1 | R-CNN(已过时) | |
| 2 | Fast R-CNN(已过时) | |
| 3 | Faster R-CNN | GitHub - kbardool/Keras-frcnn: Keras Implementation of Faster R-CNN |
| 4 | YOLO | https://github.com/ultralytics/yolov5 |
| 5 | SSD | https://github.com/lufficc/SSD |
| 6 | RetinaNet | https://github.com/fizyr/keras-retinanet |
| 7 | Mask R-CNN | https://github.com/SanmathiK/PedNet |
| 8 | EfficientDet | GitHub - xuannianz/EfficientDet: EfficientDet (Scalable and Efficient Object Detection) implementation in Keras and Tensorflow |
| 9 | CenterNet | https://github.com/xingyizhou/CenterNet |

3-目标分割任务模型

|--------|----------|----------------|-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| 序号 | 分割类型 | 模型 | ipynb模型的github链接 |
| 1 | 语义分割 | FCN | GitHub - wkentaro/pytorch-fcn: PyTorch Implementation of Fully Convolutional Networks. (Training code to reproduce the original result is available.) |
| 2 | 语义分割 | U-Net | GitHub - yingkaisha/keras-unet-collection: The Tensorflow, Keras implementation of U-net, V-net, U-net++, UNET 3+, Attention U-net, R2U-net, ResUnet-a, U^2-Net, TransUNET, and Swin-UNET with optional ImageNet-trained backbones. |
| 3 | 语义分割 | DeepLab | GitHub - fregu856/deeplabv3: PyTorch implementation of DeepLabV3, trained on the Cityscapes dataset. |
| 4 | 语义分割 | PSPNet | GitHub - Lextal/pspnet-pytorch: PyTorch implementation of PSPNet segmentation network |
| 5 | 语义分割 | SegNet | GitHub - preddy5/segnet: A Deep Convolutional Encoder-Decoder Architecture for Image Segmentation |
| 6 | 语义分割 | HRNet | GitHub - HRNet/HRNet-Image-Classification: Train the HRNet model on ImageNet |
| 7 | 实例分割 | Mask R-CNN | https://github.com/saikoneru/Instance-Segementation |
| 8 | 实例分割 | PANet | https://github.com/kaixin96/PANet |
| 9 | 实例分割 | YOLACT | https://github.com/dbolya/yolact |
| 10 | 实例分割 | SOLO | https://github.com/iambankaratharva/SOLO-Instance-Segmentation |
| 11 | 实例分割 | PointRend | https://github.com/zsef123/PointRend-PyTorch |

4-关键点检测任务模型

|--------|----------|-----------------------------------------------|-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| 序号 | 检测目标 | 模型 | ipynb模型的github链接 |
| 1 | 人脸 | Dlib | GitHub - davisking/dlib: A toolkit for making real world machine learning and data analysis applications in C++ |
| 2 | 人脸 | MTCNN | GitHub - ipazc/mtcnn: MTCNN face detection implementation for TensorFlow, as a PIP package. |
| 3 | 人脸 | FaceBoxes | GitHub - zisianw/FaceBoxes.PyTorch: A PyTorch Implementation of FaceBoxes |
| 4 | 人脸 | PRNet | GitHub - yfeng95/PRNet: Joint 3D Face Reconstruction and Dense Alignment with Position Map Regression Network (ECCV 2018) |
| 5 | 人体 | OpenPose | GitHub - Hzzone/pytorch-openpose: pytorch implementation of openpose including Hand and Body Pose Estimation. |
| 6 | 人体 | HRNet | https://github.com/HRNet/HigherHRNet-Human-Pose-Estimation |
| 7 | 人体 | CPM | GitHub - PanZiqiAI/CPM-Clothes-Keypoints-Detection: Convolutional Pose Machine implemented for clothes key points detection. |
| 8 | 人体 | Mask R-CNN with Keypoint Detection Branch | GitHub - chrispolo/Keypoints-of-humanpose-with-Mask-R-CNN: Use the Mask RCNN for the human pose estimation |
| 9 | 人体 | AlphaPose | GitHub - Amanbhandula/AlphaPose: AlphaPose Implementation in Pytorch along with the pre-trained weights |
| 10 | 人体 | MoveNet | GitHub - fire717/movenet.pytorch: A Pytorch implementation of MoveNet from Google. Include training code and pre-trained model. |

需要说明,上表中模型名称并不单指某个模型,而是一类模型统称,如YOLO模型实际包括了yolov1~yolov10的10个系列。

相关推荐
聚客AI9 小时前
🌟大模型为什么产生幻觉?预训练到推理的漏洞全揭秘
人工智能·llm·掘金·日新计划
Juchecar9 小时前
一文讲清 nn.Sequential 等容器类
人工智能
阿里云云原生9 小时前
如何快速看懂「祖传项目」?Qoder 强势推出新利器
人工智能
美团技术团队10 小时前
LongCat-Flash:如何使用 SGLang 部署美团 Agentic 模型
人工智能·算法
程序员小袁11 小时前
基于C-MTEB/CMedQAv2-rerankingv的Qwen3-1.7b模型微调-demo
人工智能
飞哥数智坊12 小时前
AI 编程一年多,我终于明白:比技巧更重要的,是熟练度
人工智能·ai编程
新智元13 小时前
收手吧 GPT-5-Codex,外面全是 AI 编程智能体!
人工智能·openai
IT_陈寒13 小时前
Java 性能优化:5个被低估的JVM参数让你的应用吞吐量提升50%
前端·人工智能·后端
阿里云云原生13 小时前
阿里云基础设施 AI Tech Day AI 原生,智构未来——AI 原生架构与企业实践专场
人工智能
Memene摸鱼日报14 小时前
「Memene 摸鱼日报 2025.9.16」OpenAI 推出 GPT-5-Codex 编程模型,xAI 发布 Grok 4 Fast
人工智能·aigc