VIP20
本课程属于：005-Artificial Intelligence and Machine Learning Engineering English Technical Courses · 共 20 门相关课程，组合学习更完整。
查看所属VIP全部课程 →
← 返回所属目录 咨询 / 购买

Few-Shot and Transfer Learning for Industrial Inspection

30 Chapters
View Course Samples Online
01
Industrial inspection landscape
defect types, quality control workflows, and why few-shot and transfer learning matter.
Industrialinspection
02
Few-shot learning fundamentals
N-way K-shot episodes, support and query sets, and evaluation protocols for inspection tasks.
Few-shotepisodes
03
Transfer learning fundamentals
pretraining, fine-tuning, feature extraction, and domain adaptation for industrial imagery.
Transferpretraining
04
Data challenges in industrial inspection
class imbalance, rare defects, labeling cost, and dataset construction strategies.
challengesindustrial
05
Pretrained backbone selection
CNN and Vision Transformer architectures for defect detection and classification.
Pretrainedbackbone
06
Fine-tuning strategies
full fine-tuning, linear probing, layer freezing, and discriminative learning rates.
Fine-tuningstrategies
07
Metric-based few-shot methods
Prototypical Networks, Matching Networks, and Relation Networks for defect recognition.
Metric-basedfew-shot
08
Optimization-based few-shot methods
MAML, Reptile, and gradient-based adaptation for inspection models.
Optimization-basedfew-shot
09
Data augmentation for few-shot inspection
geometric, photometric, and generative augmentation techniques.
augmentationfew-shot
10
Self-supervised pretraining
contrastive learning, masked image modeling, and their role in industrial inspection.
Self-supervisedpretraining
11
Domain adaptation and transfer across inspection domains
cross-product, cross-sensor, and cross-factory generalization.
adaptationtransfer
12
Anomaly detection with few normal samples
one-class classification, autoencoders, and memory banks.
detectionone-class
13
Few-shot object detection for surface defects
region proposals, metric learning, and fine-tuning detectors.
Few-shotdetection
14
Few-shot semantic segmentation for pixel-level defect localization
prototype alignment and few-shot segmentation heads.
Few-shotsemantic
15
Handling extreme class imbalance
reweighting, resampling, and focal loss in few-shot inspection.
Handlingimbalance
16
Uncertainty quantification and confidence calibration in few-shot inspection models.
Uncertainty quantification and confidence calibration in few-shot inspection models.
Uncertaintyquantification
17
Evaluation metrics for industrial inspection
precision, recall, F1, AUROC, and defect-level metrics.
Evaluationindustrial
18
Cross-validation and benchmark design for few-shot inspection
episodic evaluation and realistic splits.
Cross-validationbenchmark
19
Case study
few-shot defect classification in semiconductor manufacturing.
few-shotclassification
20
Case study
transfer learning for PCB and electronic component inspection.
transferPCB
21
Case study
few-shot surface defect detection in steel and metal production.
few-shotdetection
22
Case study
few-shot fabric and textile defect inspection.
few-shotinspection
23
Case study
few-shot anomaly detection in pharmaceutical and food packaging.
few-shotdetection
24
Case study
transfer learning for automotive paint and body inspection.
transferautomotive
25
Deployment considerations
model compression, quantization, and edge inference for inspection systems.
Deploymentconsiderations
26
MLOps for few-shot inspection
data pipelines, model versioning, and continuous learning.
MLOpsfew-shot
27
Human-in-the-loop and active learning for few-shot industrial inspection.
Human-in-the-loop and active learning for few-shot industrial inspection.
Human-in-the-loopfew-shot
28
Explainability and interpretability for few-shot inspection models in quality control.
Explainability and interpretability for few-shot inspection models in quality control.
Explainabilityinterpretability
29
Emerging trends
multimodal few-shot learning, foundation models, and zero-shot defect detection.
Emergingmultimodal
30
Capstone project
end-to-end few-shot and transfer learning pipeline for a real industrial inspection scenario.
Capstoneend-to-end