Fast eye localization without a face model using inner product detectors

A joint cascaded framework for simultaneous eye detection and eye state estimation

“…In [6], the authors propose a method for eye localization based on an ensemble of randomized regression trees, which are trained by using the pixel intensity differences around pupils. Araujo et al [7] describe an Inner Product Detector for eye localization based on the correlation filters. Zhang et al [8] use local linear SVM for 65 eye center detection, and ASEF-based filters are applied to select the candidate centers.…”

Accurate and robust eye center localization via fully convolutional networks

“…And the green points represent the ground truth positions of eye center provided by the database. The first two rows show a selection of images of different subjects with various [11] 37.0% 64.0% --96.0% Timm and Barth [13] 82.5% 93.4% 95.2% 96.4% 98.0% Cai et al [14] 84.1% 95.6% --99.8% Xia et al [62] 87.1% 98.7% --99.9% Valenti et al [12] 84.1% 90.9% 93.8% 97.0% 98.5% Soelistio et al [15] 80.8% 95.2% 97.8% 98.9% 99.4% Leo et al [17] 80.7% 87.3% 88.8% 90.9% -Leo et al [64] 78.0% 86.0% --90.0% Asadifard et al [16] 47.0% 86.0% 89.0% 93.0% 96.0% Araujo et al [19] 88.3% 92.7% 94.5% 96.3% 98.9% Niu et al [25] 75.0% 93.0% 95.8% 96.4% 97.0% Chen et al [29] -89.7% --95.7% Jesorsky et al [27] 38.0% 78.8% 84.7% 87.2% 91.8% Gou et al [32] 89.2% 98.0% --99.8% Behnke [31] 37.0% 86.0% 95.0% 97.5% 98.0% Markus et al [33] 89.9% 97.1% --99.7% Kim et al [26] -96.4% --98.8% Everingham et al [20] 45.87% 81.35% --91.21% Ren et al [35] 77.08% 92.25% --98.99% Campadelli et al [23] 80.7% 93.2% --99.3% Chen et al [24] 88.79% 95.2% --98.98% Chen et al [34] 87.3% 94.9% --99.2% Hamouz et al [22] 58.6% 75.0% 80.8% 87.6% 91.0% Kroon et al [28] 65.0% 87.0% --98.8% Cristinacce et al [30] 57.0% 96.0% 96.5% 97.0% 97.1% Hamouz et al [36] 50.0% 66.0% --70.0% Turkan et al [37] 18.6% 73.7% 94.2% 98.7% 99.6% Campadelli et al [38] 62.0% 85.2% 87.6% 91.6% 96.1% Valenti et al [39] 86.1% 91.7% --97.9% Zhang et al [47] 85.7% 93.7% --99.2% Gou et al [49] 91.2% 99.4% 99.6% -99.8% Gou et al [50] 92.3% 99.1% 99.7% --George et al [51] 85.1% 94.3% 96.7% 98.1% -Choi et al [52] 91 poses, facial expressions, occlusions and lighting co...…”

“…A method proposed by Leo et al [17] used the local variability of the appearance and image intensities to determine the eye center. Araujo et al [19] described an Inner Product Detector for eye localization based on correlation filters. The appearance-based methods have achieved good performance, but under some challenging scenarios like poor illumination they are not robust and accurate enough.…”

Real Time Eye Detector with Cascaded Convolutional Neural Networks

“…In addition, most of the existing methods rely on the accuracy of the face detector and are hard to apply in practice. In our experimental environment, the eye regions and centers were located in approximately 9 ms, wherein approximately 2 ms was spent in proposing the eye candidate feature points while 2 ms was spent in calculating the accurate region of the left eye or right eye (1st set of CNNs), and 30 ms was [5] 86.1% 91.7% 97.9% * Timm and Barth [7] 82.5% 93.4% 99.7% * Amos et al [20] 84.1% 90.2% 96.1% 500 Araujo et al [9] 88.3% 92.7% 98.9% 83 Asadifard and Shanbezadeh [25] 47.0% 86.0% 96.0% 45 Markuš et al [6] 85.7% 95.3% 99.7% 28 Leo et al 2013 [12] 78.0% 86.0% 90.0% 330 Leo et al 2014 [13] 80.6% 87.3% 94.0% 330 Gou et al [15] 91 spent in locating the center of the eye (2nd CNNs). The frame rate is around 30-60 fps for most eye detection tasks.…”

“…The RANSAC [8] method was used to create an elliptic equation to fit the pupil center. Araujo et al [9] described an Inner Product Detector for eye localization based on correlation filters. The traditional eye detectors sometimes can achieve good results, but they easily fail when there is a change in the external light or face occlusion.…”