Towards an Efficient Segmentation Algorithm for Near-Infrared Eyes Images

Abstract

Semantic segmentation has been widely used for several applications, including the detection of eye structures. This is used in tasks such as eye-tracking and gaze estimation, which are useful techniques for human-computer interfaces, salience detection, and Virtual reality (VR), amongst others. Most of the state of the art techniques achieve high accuracy but with a considerable number of parameters. This paper explores alternatives to improve the efficiency of the state of the art method, namely DenseNet Tiramisu, when applied to NIR image segmentation. This task is not trivial; the reduction of block and layers also affects the number of feature maps. The growth rate (k) of the feature maps regulates how much new information each layer contributes to the global state, therefore the trade-off amongst grown rate (k), IOU, and the number of layers needs to be carefully studied. The main goal is to achieve a lightweight and efficient network with fewer parameters than traditional architectures in order to be used for mobile device applications. As a result, a DenseNet with only three blocks and ten layers is proposed (DenseNet10). Experiments show that this network achieved higher IOU rates when comparing with Encoder-Decoder, DensetNet56-67-103, MaskRCNN, and DeeplabV3+ models in the Facebook database. Furthermore, this method reached 8th place in The Facebook semantic segmentation challenge with 0.94293 mean IOU and 202.084 parameters with a final score of 0.97147. This score is only 0,001 lower than the first place in the competition. The sclera was identified as the more challenging structure to be segmented.

Towards an Efficient Semantic Segmentation Method of ID Cards for Verification Systems

“…The method obtained stateof-the-art results for scene parsing and semantic segmentation tasks. The work of Valenzuela et al [31] proposed a method for semantic segmentation of NIR eye images, where a lightweight CNN architecture named DenseNet10, with only three blocks and ten layers is developed. The trade-off amongst grown rate (k), IoU, and the number of layers was carefully explored.…”

Semantic Segmentation of Periocular Near-Infra-Red Eye Images Under Alcohol Effects

“…Semantic segmentation is the process of identifying each object in an image. This process is performed pixel by pixel to evaluate and assign a label to each pixel [19]. State-of-the-art algorithms such as semantic segmentation have been mainly trained to localize very complex objects from cities such as cars, buildings, and people and a few in biometric gaze applications.…”

“…Valenzuela et al [19] proposed an efficient DenseNet based on DenseNet56 and compared several implementation in number of parameters, scores and complexity using DeepLabv3, UNet, Mask-RCNN, DenseNet-56, DenseNet101 and, DenseNet10. These models were trained using the OpenEDS database 1 .…”

Semantic Segmentation of Periocular Near-Infra-Red Eye Images Under Alcohol Effects

“…Semantic segmentation is the process of identify each part/subject in an image. This process is performed pixel by pixel in such a way it is owned by the object contained in the image [28].…”

“…Valenzuela et al [28] proposed an efficient DenseNet based on DenseNet56 and compared several implementation in number of parameters, scores and complexity using DeepLabV3, UNet, Mask-RCNN, DenseNet-56, DenseNet101 and, DenseNet10. These models were trained using the OpenEDS database 1 .…”