On 3rd and 4th of April I attended a workshop on Artificial Intelligence over Amazon Web Services, hosted by LinuxWorld Informatics Pvt Ltd. In this workshop we learnt why we need an AI platform provided to us by CSPs like AWS. And we also learnt the capabilities these AI models can provide to us.
Amazon Rekognition service is an object detection as well as object recognition model provider. It is very highly accurate and also Amazon provide us with SDK via which a user can integrate those models into their own apps. Confidence score is a percentage score for which the AI thinks the object detected is a certain object. If the score is high it means that the model thinks that the prediction done is very accurate. If the score is low then the model is uncertain about the prediction that it made.
To use AWS Rekognition we have two ways
1. Use the web app provided by AWS.
2. Use the AWS and use code to use Rekognition.
We can use boto3 library in python to for contacting to AWS from code.
Bounding box gives the coordinates of the object’s corner that Rekognition recognizes and through it we can find what is the object that Rek recognized and where is it located.
We can use cv2 library to connect to webcam in python.
Labels in Rek is the objects label i.e. name of the object, which Rek thinks the object could be inside the bounding box for that label.
Amazon Polly helps to covert text to speech that too in a very natural manner that sounds more human like than robot like. Polly is also capable of changing the way that it speaks, for example we can make it whisper, shout, say something slowly, make it emphasize on some point etc. This capabilities of Polly are achieved through the use of SSML.
Amazon Lex is an AWS service that has the capability to work like a chat bot i.e. it is built over NLP to be able to recognize human language patterns and processes those statement and converts them into intents and then intents can be used to trigger some action. NLP is short for natural language processing. Intents in Lex are the intention which NLP is supposed to look for i.e. they are the patterns in the provided chat or speech that a user may like to be detected to trigger some action.
Amazon Kendra is a local search engine, that can search something from local files. Amazon Kendra also behind the scene uses NLP.
Amazon Code Guru is a code profiling and reviewing service which can be used to optimize a code.
Amazon Comprehend is used to comprehend written data in image form and tries to find key phrases and words. It can also understand sentiments of statements.
Amazon Textract helps to extract a hard copy of some written form into a soft copy of that form.
Amazon translate is used to translate something from one language to other. This tool can be automated to provide cross-lingual communication.
Amazon Personalize is used to create personalized recommendation engine.
Amazon Fraud Detector can detect frauds very quickly and with high precision and thus can help large scale companies to be fraud free.