Amazon Rekognition is a (paid) service that is able to identify objects, people, text, scenes, and activities in a picture. We want to produce a free alternative.

Last year we had a number of students working on this in parallel, which in itself was an experiment - we wanted to see if having several students in parallel try to do their own implementation would be better than just choosing the one we thought would best.

This year we are only going to select one. The first task is to get the best from last year (so you have to do some analysis of the existing implementations) and come up with ways to merge (or rewrite) and improve. Therefore the chosen student must do some homework in order to come up with a good proposal. Also check the current status of Rekognition. What new features would be great to have? Which features is Rekognition actually missing and we could add ourselves to make something better than Rekognition?

This is the description to last year's project.

And here's the links to all the projects (yes we know one of them is offline at this time):

https://medium.com/@amkr/final-work-submission-gsoc19-ccextractor-development-40a2b6c6a946
https://github.com/pymit/Rekognition

https://gitlab.com/drcpmkeyi/poor-man-rekognition

https://medium.com/@sziraqui/not-the-normal-gsoc-journey-d51a6167a3a6

http://fedoskin.org/2019/08/24/gsoc-2019-poor-mans-rekognition/

https://medium.com/@amkr/final-work-submission-gsoc19-ccextractor-development-40a2b6c6a946

Qualification tasks

- An important aspect of your proposal should be the report on which of the last year's project is best or what part of each of them are best and should be reused in your projects.
- Take a look at this page for more tasks.

  • public/gsoc/poormanrekognition2.txt
  • Last modified: 2020/03/01 18:10
  • by thealphadollar