Face Detection CAPTCHA was our first image-based CAPTCHA.
In this CAPTCHA, black-and-white photographs of human faces, animal faces, and cartoon characters are visually distorted using a single image distortion. The distorted images are then embedded in a background of large multi-shaded blocks.
To solve this CAPTCHA, users must tap or click on all the human face photographs.
In testing, humans were able to successfully solve Face Detection CAPTCHA in 73.45% to 79.03% of attempts. 5.82% to 10.95% of computer attacks were successful, depending on the image distortion used. 
- B. M. Powell, A. C. Day, R. Singh, M. Vatsa, and A. Noore, “Image-based face detection CAPTCHA for improved security,” International Journal of Multimedia Intelligence and Security, vol. 1, no. 3, pp. 269–284, 2010. PDF Publisher's Website