如何在三天内选择、训练、优化和部署边缘视觉 AI 模型.pdf

1、How to Select,Train,Optimize and Deploy Edge Vision AI Models in Three Days Steven KimChief Executive OfficerNota America Inc.2023 Nota Inc.Before We Start2 When a new hardware is released in the market,customers want to run the latest AI model on the device Example Customers:Want to run YOLOv5 on a

2、 device to maximize performance(latency,accuracy,and power consumption)Device:Only supports YOLOv2 and 3 In order to close this gap,hardware-aware optimization is a must!Hardware Meets AI Software3 2023 Nota Inc.2023 Nota Inc.NetsPresso(Not The Coffee Maker)4 Expectation:How do we build the perfect

3、model for the next 6 months?Once trained,the model is completeExpectation vs Reality 2023 Nota Inc.Reality:AI model development does not end after a project AI model deployment is NOT a one-off process AI model deployment is a CONTINUAL process5 Question from AI engineers:How do we build a pipeline

4、to update and deploy a model as fast as possible?Thus,Nota AI built its own pipeline,NetsPressoWhy We Built Our Pipeline 2023 Nota Inc.6 Development process is repetitive and resource-intensivePain Point:Traditional AI Model Development Process 2023 Nota Inc.7 Once upon a time,Nota AI was started as

5、 a smart mobile keyboard app companyThe Beginning:Nota AI 2023 Nota Inc.8 Development process is even more repetitive!Pain Point:HW-Aware AI Model Development Process 2023 Nota Inc.Manual process9Our Solution:NetsPresso 2023 Nota Inc.Vision AI Model Development ProcessWithout NetsPresso:6 to 12 week

6、sWith NetsPresso:2 to 3 days10ResultsIntrusion detectionFire detectionFall DetectionSafety HelmetDetectionPothole&crack detectionHand DetectionPore DetectionPeople countingRailroad and Obstacle Detection Better performance&more diverse applications in a short time11 2023 Nota Inc.2023 Nota Inc.Case