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

Edge AI Model Optimization and Industrial Deployment

30 Chapters
View Course Samples Online
01
Edge AI deployment landscape
hardware targets, latency and power constraints, and industrial use cases.
deploymentlandscape
02
Model optimization pipeline
profiling, quantization, pruning, and knowledge distillation for edge devices.
optimizationpipeline
03
Quantization fundamentals
post-training quantization, quantization-aware training, and calibration strategies.
Quantizationpost-training
04
Pruning and sparsity
structured vs. unstructured pruning, magnitude and movement pruning, and hardware-aware sparsity.
sparsitystructured
05
Knowledge distillation for edge models
teacher-student architectures, feature distillation, and task-specific distillation.
distillationteacher-student
06
Neural architecture search for edge
hardware-aware NAS, latency predictors, and multi-objective optimization.
architecturehardware-aware
07
Efficient convolution and attention
depthwise separable convolutions, grouped convolutions, and linear attention variants.
Efficientconvolution
08
Model compression for vision transformers
token pruning, patch merging, and efficient attention for edge inference.
compressiontransformers
09
Operator fusion and graph optimization
TensorRT, OpenVINO, and TFLite converter internals.
Operatoroptimization
10
Runtime and inference engines
ONNX Runtime, TensorRT, OpenVINO, and TVM for edge deployment.
inferenceONNX
11
Hardware acceleration for edge AI
NPUs, DSPs, GPUs, and FPGAs with vendor-specific toolchains.
Hardwareacceleration
12
Memory and bandwidth optimization
activation recomputation, weight sharing, and in-memory computing.
bandwidthoptimization
13
Edge training and on-device learning
federated learning, split learning, and continual learning under resource constraints.
trainingon-device
14
Model serving on edge gateways
containerization, orchestration, and OTA model updates.
gatewayscontainerization
15
Benchmarking and performance analysis
latency, throughput, energy, and accuracy trade-offs on edge devices.
Benchmarkingperformance
16
Industrial deployment case study
smart manufacturing with real-time defect detection on edge devices.
Industrialdeployment
17
Industrial deployment case study
autonomous mobile robots with optimized perception and navigation models.
Industrialdeployment
18
Industrial deployment case study
predictive maintenance on edge gateways for rotating machinery.
Industrialdeployment
19
Industrial deployment case study
smart retail with edge-based customer analytics and inventory monitoring.
Industrialdeployment
20
Industrial deployment case study
intelligent transportation with edge-optimized traffic sign recognition and pedestrian detection.
Industrialdeployment
21
Security and privacy in edge AI
model encryption, secure enclaves, and adversarial robustness.
Securityencryption
22
Reliability and fault tolerance
graceful degradation, model fallback, and watchdog mechanisms.
Reliabilitytolerance
23
Energy-efficient inference
dynamic voltage and frequency scaling, power capping, and energy-aware scheduling.
Energy-efficientinference
24
Multi-model and multi-task edge deployment
model cascades, early exit, and conditional computation.
Multi-modelmulti-task
25
Edge AI for video analytics
frame sampling, region-of-interest processing, and temporal fusion.
analyticssampling
26
Edge AI for audio and speech
keyword spotting, noise suppression, and on-device speech recognition.
spottingsuppression
27
Edge AI for time-series and sensor data
anomaly detection, forecasting, and event classification.
time-seriesdetection
28
Deployment toolchains and MLOps for edge
model conversion, versioning, monitoring, and rollback.
Deploymenttoolchains
29
Performance tuning and debugging
profiling tools, bottleneck identification, and optimization iterations.
Performancedebugging
30
Future trends and challenges
tinyML, neuromorphic computing, and sustainable edge AI deployment.
challengestinyML