Telling computers and humans apart automatically using activity recognition
E. R. Vimina, Alba Urm Areekal · 2009
This paper proposes a new image based CAPTCHA test, activity recognition CAPTCHA. In this test the user is presented with a set of distorted images depicting a randomly chosen activity. The user has to recognize the common activity associated with the images and annotate it from a given list of activities to pass the test. The user studies indicate that this CAPTCHA can be solved with 99.04% average pass rate while that of ESP PIX CAPTCHA is 85.44% and SQUIGL PIX is 67.84%. Average time taken to pass the test is less than 10 seconds.