Which is the FASTEST route to index the assets?
Use Amazon Rekognition, Amazon Comprehend, and Amazon Transcribe to tag data into distinct categories/classes.
Create a set of Amazon Mechanical Turk Human Intelligence Tasks to label all footage.
Use Amazon Transcribe to convert speech to text. Use the Amazon SageMaker Neural Topic Model (NTM) and Object Detection algorithms to tag data into distinct categories/classes.
Use the AWS Deep Learning AMI and Amazon EC2 GPU instances to create custom models for audio transcription and topic modeling, and use object detection to tag data into distinct categories/classes.
Explanations:
This option leverages AWS managed services that are designed for fast deployment and ease of use. Amazon Rekognition can automatically tag images, Amazon Comprehend can analyze and extract insights from text, and Amazon Transcribe can convert audio to text quickly. This approach requires minimal ML expertise and allows rapid indexing of diverse asset types using pre-trained models.
While Amazon Mechanical Turk can provide human intelligence for labeling tasks, it is a slower and more manual process compared to automated machine learning solutions. Labeling all footage would take significant time and effort, making it less suitable for a rapid indexing solution.
This option combines Amazon Transcribe for speech-to-text conversion and suggests using NTM and Object Detection algorithms. However, implementing these algorithms requires more expertise in machine learning and potentially longer development time, which contradicts the goal of fast indexing. Furthermore, NTM may not be directly available as a managed service like the others.
Using the AWS Deep Learning AMI and Amazon EC2 GPU instances to create custom models is likely the slowest route as it involves building and training models from scratch. This requires significant machine learning expertise and is not optimal for a media company looking for a quick and efficient indexing solution.