Image Processing

    OpenCL has become the favored framework for emerging heterogeneous devices and FPGAs, owing to its versatility and portability.

    However, OpenCL-based math libraries still face challenges in fully leveraging device performance.

    When deploying high-performance arithmetic applications on these devices, the most important hot function is General Matrix-matrix Multiplication (GEMM).

    This study presents a meticulously optimized OpenCL GEMM kernel.

Categories:
52 Views

The Deepfake-Synthetic-20K dataset significantly contributes to digital forensics and deepfake detection research. It comprises 20,000 high-resolution, synthetic human face images generated using the advanced StyleGAN-2 architecture. This dataset is designed to support the development and evaluation of machine-learning models that can differentiate between real and artificially synthesized human faces. Each image in the dataset has been meticulously crafted to ensure a diverse representation of age, gender, and ethnicity, reflecting the variability seen in global human populations.

Categories:
396 Views

DIRS24.v1 presents a dataset captured in campus environment. These images are curated suitably for the utilization in developing perception modules. These modules can be very well employed in Advanced Driver Assistance Systems (ADAS). The images of dataset are annotated in diversified formats such as COCO-MMDetection, Pascal-VOC, TensorFlow, YOLOv7-PyTorch, YOLOv8-Oriented Bounding Box, and YOLOv9.

 

Categories:
344 Views

The foundation of detection relies on the surface micro-defect images of KDP, and the effectiveness of the detection model depends on the quality of these images. Higher quality images can pinpoint the shape details and boundary features of defects, thereby enhancing the overall detection capability.

Categories:
97 Views

As an artificial structure, tailings ponds exhibit regular geometric shapes and relatively straight dams in HRRSIs. Because the typical tailings dam is composed of an initial dam and successive accumulation dams, the tailings dam structure presents obvious linear stripe characteristics. The initial dam, constructed using sand, gravel, or concrete, has a bright color, while the color of the accumulation dam varies based on factors such as particle size, soil coverage, and vegetation restoration.

Categories:
65 Views

This dataset is used to compressive sensing scenes and assist users in training the optimal sparsity threshold. The input features are image compression ratio, image size, and sparsity, while the label field is the optimal sparsity threshold. The test images are from the USC-SIPI dataset. 

Categories:
27 Views

The Colour-Rendered Bosphorus Projections (CRBP) Face Dataset represents an innovative advancement in facial recognition and computer vision technologies. This dataset uniquely combines the precision of 3D face modelling with the detailed visual cues of 2D imagery, creating a multifaceted resource for various research activities. Derived from the acclaimed Bosphorus 3D Face Database, the CRBP dataset introduces colour-rendered projections to enrich the original dataset.

Categories:
245 Views

This work presents a new labeled dataset of videos with native and professional interpreters articulating words and expressions in Libras (Brazilian Sign Language). We used a methodology based on related studies, the support of the team of articulators, and the existing datasets in the literature.

Categories:
355 Views

The "MANUU: Handwritten Urdu OCR Dataset" is an extensive and meticulously curated collection to advance OCR (Optical Character Recognition) for handwritten Urdu letters, digits, and words. The compilation of the dataset has been conducted methodically, ensuring that it encompasses a wide variety of handwritten instances. This comprehensive collection enables the construction and assessment of strong models for Optical Character Recognition (OCR) systems specifically designed for the complexities of the Urdu script.

Categories:
256 Views

Video super-resolution (SR) has important real world applications such as enhancing viewing experiences of legacy low-resolution videos on high resolution display devices. However, there are no visual quality assessment (VQA) models specifically designed for evaluating SR videos while such models are crucially important both for advancing video SR algorithms and for viewing quality assurance. Therefore, we establish a super-resolution video quality assessment database (VSR-QAD) for implementing super-resolution video quality assessment.

Categories:
105 Views

Pages