Sayan Sarcar

Zhenxin Wang, Sayan Sarcar, Jingxin Liu et al.

Image segmentation needs both local boundary position information and global object context information. The performance of the recent state-of-the-art method, fully convolutional networks, reaches a bottleneck due to the neural network limit after balancing between the two types of information simultaneously in an end-to-end training style. To overcome this problem, we divide the semantic image segmentation into temporal subtasks. First, we find a possible pixel position of some object boundary; then trace the boundary at steps within a limited length until the whole object is outlined. We present the first deep reinforcement learning approach to semantic image segmentation, called DeepOutline, which outperforms other algorithms in Coco detection leaderboard in the middle and large size person category in Coco val2017 dataset. Meanwhile, it provides an insight into a divide and conquer way by reinforcement learning on computer vision problems.

Sayan Sarcar, Ahmed Sabbir Arif, Ali Mazalek

With the intention of bringing uniformity to Bengali text entry research, here we present a new approach for calculating the most popular English text entry evaluation metrics for Bengali. To demonstrate our approach, we conducted a user study where we evaluated four popular Bengali text entry techniques.

Sayan Sarcar

Dwelling is an essential task to be performed to select keys from an on-screen keyboard present in the eye typing interface. This selection task can be performed by fixing eye gaze on a key for a prolonged time. Spending sufficient amount of time on each key effectively decreases the overall eye typing rate. To address the problem, researchers proposed mechanisms, which diminish the dwell time. We conducted a within-subject usability evaluation of four dwell-free eye typing techniques. The results of first-time usability study, longitudinal study and subjective evaluation conducted with 15 participants confirm the superiority of controlled eye movement based advanced eye typing method (Adv-EyeK) than the other three techniques.

Pawan Patidar, Himanshu Raghuvanshi, Sayan Sarcar

Pie menus are suggested as powerful tool for eye gaze based text entry among various interfaces developed so far. If pie menus are used with multiple depth layers then multiple saccades are required per selection of item, which is inefficient because it consumes more time. Also dwell time selection method is limited in performance because higher dwell time suffers from inefficiency while lower one from inaccuracy. To overcome problems with multiple depth layers and dwell time, we designed Quickpie, an interface for eye gaze based text entry with only one depth layer of pie menu and selection border as selection method instead of dwell time. We investigated various parameters like number of slices in pie menu, width characters and safe region, enlarged angle of slice and selection methods to achieve better performance. Our experiment results indicates that six number of slices with width of characters area 120 px performs better as compared to other designs.