Using cascade CNN-LSTM-FCNs to identify AI-altered video based on eye state sequence.

Deep learning is notably successful in data analysis, computer vision, and human control. Nevertheless, this approach has inevitably allowed the development of DeepFake video sequences and images that could be altered so that the changes are not easily or explicitly detectable. Such alterations have been recently used to spread false news or disinformation. This study aims to identify Deepfaked videos and images and alert viewers to the possible falsity of the information. The current work presented a novel means of revealing fake face videos by cascading the convolution network with recurrent neural networks and fully connected network (FCN) models. The system detection approach utilizes the eye-blinking state in temporal video frames. Notwithstanding, it is deemed challenging to precisely depict (i) artificiality in fake videos and (ii) spatial information within the individual frame through this physiological signal. Spatial features were extracted using the VGG16 network and trained with the ImageNet dataset. The temporal features were then extracted in every 20 sequences through the LSTM network. On another note, the pre-processed eye-blinking state served as a probability to generate a novel BPD dataset. This newly-acquired dataset was fed to three models for training purposes with each entailing four, three, and six hidden layers, respectively. Every model constitutes a unique architecture and specific dropout value. Resultantly, the model optimally and accurately identified tampered videos within the dataset. The study model was assessed using the current BPD dataset based on one of the most complex datasets (FaceForensic++) with 90.8% accuracy. Such precision was successfully maintained in datasets that were not used in the training process. The training process was also accelerated by lowering the computation prerequisites.

Dorsal Hand Vein Image Enhancement Using Fusion of CLAHE and Fuzzy Adaptive Gamma.

Enhancement of captured hand vein images is essential for a number of purposes, such as accurate biometric identification and ease of medical intravenous access. This paper presents an improved hand vein image enhancement technique based on weighted average fusion of contrast limited adaptive histogram equalization (CLAHE) and fuzzy adaptive gamma (FAG). The proposed technique is applied using three stages. Firstly, grey level intensities with CLAHE are locally applied to image pixels for contrast enhancement. Secondly, the grey level intensities are then globally transformed into membership planes and modified with FAG operator for the same purposes. Finally, the resultant images from CLAHE and FAG are fused using improved weighted averaging methods for clearer vein patterns. Then, matched filter with first-order derivative Gaussian (MF-FODG) is employed to segment vein patterns. The proposed technique was tested on self-acquired dorsal hand vein images as well as images from the SUAS databases. The performance of the proposed technique is compared with various other image enhancement techniques based on mean square error (MSE), peak signal-to-noise ratio (PSNR), and structural similarity index measurement (SSIM). The proposed enhancement technique's impact on the segmentation process has also been evaluated using sensitivity, accuracy, and dice coefficient. The experimental results show that the proposed enhancement technique can significantly enhance the hand vein patterns and improve the detection of dorsal hand veins.

Wind shear data at two different terrain types.

The measurement of this data aims to evaluate the wind shear variability at three selected sites in Malaysia. The sites are Kudat in Sabah, Kijal in Terengganu and Langkawi in Kedah. Both sites in Kudat and Kijal is located in coastal areas with few buildings or trees, while the site in Langkawi is a coastal area with many buildings or trees. The variables were measured using the sensors that mounted on the wind mast with the maximum height from 55 m to 70 m from ground level. The variables measured were wind speed, wind direction, temperature, and pressure, while the wind shear data were directly generated using the power law equation. The averaged wind shear based on measured multiple height wind speed at selected sites is larger than the 1/7 law (0.143). Also, the value of wind shear was higher in order Langkawi > Kudat > Kijal. Ultimately, the wind shear data are essential and can be reused in the wind energy potential study, especially for data extrapolation to desired wind turbine hub height.