Denoising videos in real-time is critical many applications, including robotics and medicine, where varying-light conditions, miniaturized sensors, optics can substantially compromise image quality. This work proposes the first video denoising method based on a deep neural network that achieves state-of-the-art performance dynamic scenes while running VGA resolution with no frame latency. The backbone of our novel, remarkably simple, temporal cascaded blocks forward block output propagation....

Sperm tail morphology and motility have been demonstrated to be important factors in determining sperm quality for vitro fertilization. However, many existing computer-aided analysis systems leave the out of analysis, as detecting a few pixels is challenging. Moreover, some publicly available datasets classifying morphological defects contain images limited only head. This study focuses on segmentation full sperm, which consists head parts, appear alone groups. We re-purpose Feature Pyramid...

Abstract Objective . Quantitative evaluation protocols are critical for the development of algorithms that remove artifacts from real electroencephalography (EEG) optimally. However, visually inspecting EEG to select top-performing artifact removal pipeline is infeasible while hand-crafted data allow assessing configurations only in a simulated environment. This study proposes novel, principled approach quantitatively evaluating algorithmically corrected without access ground truth...

Elongated objects have various shapes and can shift, rotate, change scale, be rigid or deform by flexing, articulating, vibrating, with examples as varied a glass bottle, robotic arm, surgical suture, finger pair, tram, guitar string. This generally makes tracking of poses elongated very challenging. We describe unified, configurable framework for the pose objects, which move in image plane extend over region. Our method strives simplicity, versatility, efficiency. The object is decomposed...

Dynamic scene video deblurring is a challenging task due to the spatially variant blur inflicted by independently moving objects and camera shakes. Recent deep learning works bypass ill-posedness of explicitly deriving kernel pixel-to-pixel mappings, which commonly enhanced larger region awareness. This difficult yet simplified scenario because noise neglected when it omnipresent in wide spectrum processing applications. Despite its relevance, problem concurrent dynamic has not been...

We propose a novel probabilistic tracking algorithm based on an elastic blob ensemble (EBE) which is applicable to track flexible objects. It outputs coarse motion cue in the form of object's location and orientation together with blobs. The main assumption that whole object does not change much between neighboring frames. A discrete solution space created current frame around blobs' positions from previous frame. Our model then promotes solutions whose orientations are close prior...

Haar-like features are ubiquitous in computer vision, e.g. for Viola and Jones face detection or local descriptors such as Speeded-Up-Robust-Features. They classically computed one pass over integral image by reading the values at feature corners. Here we present a new, general parsing formalism convolving them more efficiently. Our method is fully automatic applicable to an arbitrary set of features. The parser reduces number memory accesses which main computational bottleneck during...

Preventing early progression of epilepsy and so the severity seizures requires effective diagnosis. Epileptic transients indicate ability to develop but humans overlook such brief events in an electroencephalogram (EEG) what compromises patient treatment. Traditionally, training EEG event detection algorithms has relied on ground truth labels, obtained from consensus majority labelers. In this work, we go beyond labeler data. Our descriptor integrates signal features with one-hot encoded...

In dentistry, microscopes have become indispensable optical devices for high-quality treatment and micro-invasive surgery, especially in the field of endodontics. Recent machine vision advances enable more advanced, real-time applications including but not limited to dental video deblurring workflow analysis through relevant metadata obtained by instrument motion trajectories. To this end, proposed work addresses segmentation a Multi-task Learning fashion, leveraging spatio-temporal adaptive...

In visual tracking of surgical instruments, correlation filtering finds the best candidate with maximal peak. However, most trackers only consider capturing target appearance but not structure. this paper we propose instrument approach that integrates prior knowledge related to rotation both shaft and tool tips. To end, employ rigid parts mixtures model an instrument. The rigidly composed encode diverse, pose-specific tool. Tracking search space is confined neighbourhood position, scale,...

We propose an algorithm for exploring the entire regularization path of asymmetric-cost linear support vector machines. Empirical evidence suggests predictive power machines depends on parameters training algorithms. The algorithms paths have been proposed single-cost thereby providing complete knowledge behavior trained model over hyperparameter space. Considering problem in two-dimensional space though enables our to maintain greater flexibility dealing with special cases and sheds light...

Preventing early progression of epilepsy and so the severity seizures requires an effective diagnosis. Epileptic transients indicate ability to develop but humans overlook such brief events in electroencephalogram (EEG) what compromises patient treatment. Traditionally, training EEG event detection algorithms has relied on ground truth labels, obtained from consensus majority labelers. In this work, we go beyond labeler data. Our descriptor integrates signal features with one-hot encoded...