Vision API - pros and cons
Use Cases and Deployment Scope
we used Vision API to extract attribute values from medical devices images for internal purpose only. We have TBs of data to process in such and the response time was in within few seconds only. but for such a scale the processing time is coming out to be in months. then we had to introduce multiprocessing for speeding up.
Pros
- Object detection is good
- Get facial expressions across different faces from different part of the words
- Used OCR for different languages in india
Cons
- response Time is an issue
- Google API is not available across different platforms as an integrated feature
- Not able to add custom labels easily, the process is long for it
Return on Investment
- The overall impact is positive as it is pay as per use basis
- I believe there should be enterprise plans for large scale use cases where it tends to get very expensive
- some open source models provide similar functionality, so in cases of high cost we use those
Usability
Alternatives Considered
Amazon Rekognition, Qwen and Anthropic Claude
Other Software Used
Amazon Rekognition, Anthropic Claude, Azure OpenAI Service