AWS re:Invent 2017: Building a Predictive API using a Pre-Trained MXNet Model and AW (DEM76)
323
107.7
Amazon Web Services775 тыс
Следующее
Опубликовано 30 ноября 2017, 17:13
We will show you how to build an API that predicts geolocation based on image alone. This compute-intensive workload runs on AWS Lambda and scales automatically with a serverless architecture.
The flow of the demo at a high level is:
1) Upload pre-trained MXNet Model into S3
2) Deploy Lambda template
3) SAM creates API Gateway + Lambda
4) Use API to predict appropriate label for the image
At the function level, the Lambdas:
1) Download the model and store it as a global variable
2) Parse event, extract the image URL and download it
3) Run the image to the model and get the prediction
4) Send the prediction output via API Gateway
The flow of the demo at a high level is:
1) Upload pre-trained MXNet Model into S3
2) Deploy Lambda template
3) SAM creates API Gateway + Lambda
4) Use API to predict appropriate label for the image
At the function level, the Lambdas:
1) Download the model and store it as a global variable
2) Parse event, extract the image URL and download it
3) Run the image to the model and get the prediction
4) Send the prediction output via API Gateway
Свежие видео
1 день – 2 4071:20
4 дня – 397 8063:04:33
8 дней – 4 95723:58
Случайные видео
147 дней – 3 247 1939:00
211 дней – 423 5730:52
335 дней – 9 4050:25
04.10.23 – 46 5740:42
16.03.23 – 43 96710:24
20.10.09 – 307 2614:02
6 часов – 620:23
1 день – 3 0952:06
4 дня – 222 2720:16
4 дня – 6770:34
6 дней – 1300:18
8 дней – 1 9291:11
13 часов – 209 6295:15