Providing state of the art
algorithms for monitoring
livestock’s health status
SoundTalks
The challenge
There are multiple researchers at SoundTalks that actively develop the algorithm that detects sick pigs. This research requires as much data as possible to develop and test the algorithm. Over the years, SoundTalks gathered a lot of data resulting in huge processing times which is becoming too much for their local server to handle.
The solution
Because this development process only occurs a few times each month (and the server is idle the rest of the time), it made sense to migrate this processing to a cloud service provider. SoundTalks can rent a lot of servers at once for a very short time, which results in a more cost-effective approach compared to owning a couple of servers full-time (which also require maintenance).
Sentigrate developed a system that consisted of multiple components:
- A front-end interface for the researcher where he can start a calculation run. This interface sends multiple requests for all the audio-files that need to be processed.
- A load balancer that receives these requests and dynamically creates and maintains containers who process these audio files.
- A processing-container that contains the detection algorithm that runs on an audio file.
- Appropriate storage systems that contain the data and the processing results.
