Tag: Computer Vision

CTVSeg-UNet improved vein-pixel recall from 47.12% to 77.03% over standard U-Net while extracting more continuous tobacco leaf networks across curing ...

A multimodal SegFormer system measured live-chicken shank length through anatomical masks and geometric anchors, achieving low repeated-measurement variation but substantial ...

A multi-scale vision model estimated guidewire contact forces from planar images with a mean absolute error of 0.861 g on ...

A benchmark study combining adaptive background modeling, four-frame differencing and occlusion-aware tracking reached 92.9% [email protected] on ISSIA-CNR and 93.1% on ...

A foundation model trained on more than 40 million facial images produced tools that estimate biological age and directly predict ...

An enhanced YOLOv11 model developed for printed circuit board inspection achieved mean average precision of 96.1% at an intersection-over-union threshold ...

A hybrid deep learning framework combining Siamese networks and Vision Transformers achieves up to 95 percent accuracy in verifying parent-child ...

A new AI framework called GECST combines generative adversarial networks with classical compositional rules to restore China's fading Dunhuang murals ...

A new lightweight deep learning model called WheatNet detects and counts individual wheat seedlings in complex field images with record ...

Researchers in China developed MP-CBA-YOLO, a morphology-preserving, class-balanced object detection method that grades ginseng roots with expert-level accuracy despite scarce ...

For millions of people living with gout, the painful reality of the disease is written in lumps of crystallized uric ...

Researchers at Xinjiang University developed BGAFS-UNet, a boundary-aware deep learning architecture that improves retinal vessel segmentation fidelity while openly documenting ...

Researchers developed GDH-Det, an underwater object detection method that uses a conditional diffusion network to synthesize degraded training images and ...

A new deep learning framework detects violent incidents in real surveillance footage with 90 percent accuracy while running fast enough ...

A new spatio-temporal graph network models players and the ball as a dynamic graph, detecting key match events and tactical ...

A research team has enhanced YOLOv8-Pose with a StarNet backbone, motion-relevant attention, and a composite keypoint loss to jointly classify ...

Researchers in Pune have built a Dual GAN–Diffusion generative learning system that synthesizes realistic, diverse pomegranate disease images from field-acquired ...

Researchers have unveiled MiRAGE, a parameter-efficient AI framework that adapts its computational depth to each satellite image and uses CLIP ...

A new RGB-depth salient object detection framework combines multi-level attention-based feature fusion with a WebP compression and denoising defense that ...

A new multi-modal AI method pairs structured text descriptions generated by a large language model with image analysis to identify ...