Butterfly network: a convolutional neural network with a new architecture for multi-scale semantic segmentation of pedestrians.

The detection of multi-scale pedestrians is one of the challenging tasks in pedestrian detection applications. Moreover, the task of small-scale pedestrian detection, i.e., accurate localization of pedestrians as low-scale target objects, can help solve the issue of occluded pedestrian detection as well. In this paper, we present a fully convolutional neural network with a new architecture and an innovative, fully detailed supervision for semantic segmentation of pedestrians. The proposed network has been named butterfly network (BF-Net) because of its architecture analogous to a butterfly. The proposed BF-Net preserves the ability of simplicity so that it can process static images with a real-time image processing rate. The sub-path blocks embedded in the architecture of the proposed BF-Net provides a higher accuracy for detecting multi-scale objective targets including the small ones. The other advantage of the proposed architecture is replacing common batch normalization with conditional one. In conclusion, the experimental results of the proposed method demonstrate that the proposed network outperform the other state-of-the-art networks such as U-Net + + , U-Net3 + , Mask-RCNN, and Deeplabv3 + for the semantic segmentation of the pedestrians.

A New Algorithm for Skin Lesion Border Detection in Dermoscopy Images.

BACKGROUND: With advances in medical imaging systems, digital dermoscopy has become one of the major imaging modalities in the analysis of skin lesions. Thus, automated segmentation or border detection has a great impact on the subsequent steps of skin cancer computer-aided diagnosis using demoscopy images. Since dermoscopy images suffer from artifacts such as shading and hair, there is a need for automated and robust artifact attenuation removal and lesion border detection. METHODS: method for segmentation of dermoscopy images is proposed based on active contour. To this end, at first, a simple method for hair pixels is restored and a new scheme for shading detection is proposed. Then, particle swarm optimization (PSO) algorithm is applied to select the best coefficients for converting RGB to gray level. The obtained gray level image is then used as input for multi Otsu method which provides initial contour for border detection using active contour. Finally, Chan and Vese active contour is used for final lesion border detection. RESULTS: The method is tested on a total of 145 dermoscopic images: 79 cases with benign lesion and 75 cases with melanoma lesion. Mean accuracy, sensitivity and specificity were obtained 94%, 78.5% and 99%, respectively. CONCLUSION: Results reveal that the proposed method segments the lesion from dermoscopy images accurately.