How leikunx/ai_projects works
This repository is a collection of independent AI project examples, not one application. Its clearest end-to-end workflow is the YOLOv3 image-detection service: a client submits an image, the server invokes a detector, and the result is returned for client-side display. Other branches cover custom detector training, Mask R-CNN, face recognition, motion detection, and a ResNet parallel-model utility. Relationships are shown only where the supplied excerpts or SOURCE INDEX support them; unsampled project details are not inferred.
Main entry points: _11_yolov3_server_3.py Detection service · FaceDetection_mtcnn.py MTCNN detection · _02_demo.py Recognition demo · _03_application.py Face application
Components 31
YOLOv3 training
- Image selection
001_目标检测实践_keras版YOLOv3训练自己的数据/code/_01_select_images.py - Label validation
001_目标检测实践_keras版YOLOv3训练自己的数据/code/_02_check_labels.py - Image compression
001_目标检测实践_keras版YOLOv3训练自己的数据/code/_03_compress_images.py - Annotation lists
001_目标检测实践_keras版YOLOv3训练自己的数据/code/_04_generate_txtFile.py - Anchor clustering
001_目标检测实践_keras版YOLOv3训练自己的数据/code/kmeans.py - Weight conversion
001_目标检测实践_keras版YOLOv3训练自己的数据/code/convert.py - Model training
001_目标检测实践_keras版YOLOv3训练自己的数据/code/_05_train.py - YOLO model
001_目标检测实践_keras版YOLOv3训练自己的数据/code/yolo3/model.py - Image detection
001_目标检测实践_keras版YOLOv3训练自己的数据/code/_06_yolo.py - Batch images
001_目标检测实践_keras版YOLOv3训练自己的数据/code/_08_detect_multi_images.py - Video detection
001_目标检测实践_keras版YOLOv3训练自己的数据/code/yolo_video.py
YOLOv3 deployment
- Image client
005_模型部署_在线预测/code/_12_yolov3_client_3.py - Detection service
005_模型部署_在线预测/code/_11_yolov3_server_3.py - YOLO detector
005_模型部署_在线预测/code/_06_yolov3.py - YOLO model
005_模型部署_在线预测/code/yolo3/model.py
Mask R-CNN
- Data preparation
002_目标检测实践_keras版Mask-RCNN训练自己的数据/code/_01_select_images.py - Label checks
002_目标检测实践_keras版Mask-RCNN训练自己的数据/code/_02_check_labels.py - Image resizing
002_目标检测实践_keras版Mask-RCNN训练自己的数据/code/_03_resize_images.py - Model training
002_目标检测实践_keras版Mask-RCNN训练自己的数据/code/_05_train.py - Image inference
002_目标检测实践_keras版Mask-RCNN训练自己的数据/code/_06_test_one_image.py - Batch inference
002_目标检测实践_keras版Mask-RCNN训练自己的数据/code/_07_test_multi_images.py
Face recognition
- MTCNN detection
007_人脸识别实践_mxnet版insightFace/code/FaceDetection_mtcnn.py - Face alignment
007_人脸识别实践_mxnet版insightFace/code/_01_align_face.py - Face recognition
007_人脸识别实践_mxnet版insightFace/code/FaceRecognizer.py - Recognition variant
007_人脸识别实践_mxnet版insightFace/code/FaceRecognizer_2.py - Recognition demo
007_人脸识别实践_mxnet版insightFace/code/_02_demo.py - Face application
007_人脸识别实践_mxnet版insightFace/code/_03_application.py
Other vision examples
- Motion detection
006_检测移动物体/code - Parallel ResNet
004_复写代码_keras版ResNet/code/parallel_model.py
Other components
- Image user
- Detection result
Connections 10
- Image client → Detection service: POSTs image
005_模型部署_在线预测/code/_12_yolov3_client_3.py - Detection service → YOLO detector: invokes detector
005_模型部署_在线预测/code/_11_yolov3_server_3.py - YOLO detector → YOLO model: builds model
005_模型部署_在线预测/code/_06_yolov3.py - Detection service → Detection result: returns detections
- Image user → Image client: submits image
- Face alignment → MTCNN detection: uses detector
007_人脸识别实践_mxnet版insightFace/code/_01_align_face.py - Recognition demo → MTCNN detection: uses detector
007_人脸识别实践_mxnet版insightFace/code/_02_demo.py - Recognition demo → Face recognition: uses recognizer
007_人脸识别实践_mxnet版insightFace/code/_02_demo.py - Face application → MTCNN detection: uses detector
007_人脸识别实践_mxnet版insightFace/code/_03_application.py - Face application → Recognition variant: uses recognizer
007_人脸识别实践_mxnet版insightFace/code/_03_application.py