A5014378041- Advanced Image Processing Techniques
- Generative Adversarial Networks and Image Synthesis
- Reproductive Biology and Fertility
- Advanced Vision and Imaging
- Image Processing Techniques and Applications
- Software System Performance and Reliability
- Software Engineering Research
- Software Engineering Techniques and Practices
- Image Enhancement Techniques
- Ovarian function and disorders
- Domain Adaptation and Few-Shot Learning
- Advanced Neural Network Applications
- Multimodal Machine Learning Applications
- AI in cancer detection
- Prenatal Screening and Diagnostics
- Medical Image Segmentation Techniques
- Video Analysis and Summarization
- Advanced Bandit Algorithms Research
- Pluripotent Stem Cells Research
- Reproductive System and Pregnancy
- Advanced Data Compression Techniques
- Consumer Market Behavior and Pricing
- Assisted Reproductive Technology and Twin Pregnancy
- Recommender Systems and Techniques
- Digital Media Forensic Detection
Hebrew University of Jerusalem
2020-2023
Microsoft (Israel)
2022
In in vitro fertilization (IVF) treatments, early identification of embryos with high implantation potential is required for shortening time to pregnancy while avoiding clinical complications the newborn and mother caused by multiple pregnancies. Current classification tools are based on morphological morphokinetic parameters that manually annotated using time‐lapse video files. However, manual annotation introduces interobserver intraobserver variability provides a discrete representation...
The problem of Next Basket Recommendation (NBR) addresses the challenge recommending items for next basket a user, based on her sequence prior baskets. In this paper, we focus variation in which aim to predict repurchases, i.e. wish recommend user only she had purchased before. We coin Repurchase (NBRR). Over years, variety models have been proposed address NBR, however, NBRR has overlooked. Although being highly related problems, are often solved by same methods, repurchase recommendation...
Our objective was to design an automated deep learning model that extracts the morphokinetic events of embryos were recorded by time-lapse incubators. Using annotation, we set out characterize temporal heterogeneity preimplantation development across a large number embryos.
In this paper, we present DeepSIM, a generative model for conditional image manipulation based on single image. We find that extensive augmentation is key enabling training, and incorporate the use of thin-plate-spline (TPS) as an effective augmentation. Our network learns to map between primitive representation itself. The choice has impact ease expressiveness manipulations can be automatic (e.g. edges), manual segmentation) or hybrid such edges top segmentations. At time, our generator...
The colorization of grayscale images is a complex and subjective task with significant challenges. Despite recent progress in employing large-scale datasets deep neural networks, difficulties controllability visual quality persist. To tackle these issues, we present novel image framework that utilizes diffusion techniques granular text prompts. This integration not only produces outputs are semantically appropriate but also greatly improves the level control users have over process. Our...
Abstract The majority of human embryos, whether naturally or in vitro fertilized (IVF), do not poses the capacity to implant within uterus and reach live birth. Hence, selecting embryos with highest developmental potential is imperative for improving pregnancy rates without prolonging time pregnancy. can be assessed based on temporal profiling discrete morphokinetic events preimplantation development. However, manual annotation introduces intra- inter-observer variation time-consuming. Using...
Semantic segmentation is a key computer vision task that has been actively researched for decades. In recent years, supervised methods have reached unprecedented accuracy, however they require many pixel-level annotations every new class category which very time-consuming and expensive. Additionally, the ability of current semantic networks to handle large number categories limited. That means images containing rare are unlikely be well segmented by methods. this paper we propose novel...
We introduce LTX-Video, a transformer-based latent diffusion model that adopts holistic approach to video generation by seamlessly integrating the responsibilities of Video-VAE and denoising transformer. Unlike existing methods, which treat these components as independent, LTX-Video aims optimize their interaction for improved efficiency quality. At its core is carefully designed achieves high compression ratio 1:192, with spatiotemporal downscaling 32 x 8 pixels per token, enabled...
We present AugurOne, a novel approach for training single image generative models. Our trains an upscaling neural network using non-affine augmentations of the (single) input image, particularly including non-rigid thin plate spline warps. The extensive significantly increase in-sample distribution upsampling enabling highly variable inputs. A compact latent space is jointly learned allowing controlled synthesis. Differently from Single Image GAN, our does not require GAN and takes place in...
ABSTRACT In IVF treatments, accurate assessment of the developmental potential embryos to implant is essential for reaching reasonable pregnancy rates while shortening time-to-pregnancy. Hence, clinical guidelines recommend extended incubation blastocyst transfers, which provide better evaluation embryo potential. However, cleavage stage transfer often favored owing various considerations. To improve without incubation, we present a computational strategy forecasting future morphokinetic...
The colorization of grayscale images is a complex and subjective task with significant challenges. Despite recent progress in employing large-scale datasets deep neural networks, difficulties controllability visual quality persist. To tackle these issues, we present novel image framework that utilizes diffusion techniques granular text prompts. This integration not only produces outputs are semantically appropriate but also greatly improves the level control users have over process. Our...
In this paper, we present DeepSIM, a generative model for conditional image manipulation based on single image. We find that extensive augmentation is key enabling training, and incorporate the use of thin-plate-spline (TPS) as an effective augmentation. Our network learns to map between primitive representation itself. The choice has impact ease expressiveness manipulations can be automatic (e.g. edges), manual segmentation) or hybrid such edges top segmentations. At time, our generator...
Using multiple monitors is commonly thought to improve productivity, but this hard check experimentally. We use a survey, taken by 101 practitioners of which 80% have coded professionally for at least 2 years, assess subjective perspectives based on experience. To validity, we compare situations in developers naturally different setups-the difference between working home or the office, and how things changed when were forced work from due Covid-19 pandemic. The results indicate that using...
Using multiple monitors is commonly thought to improve productivity, but this hard check experimentally. We use a survey, taken by 101 practitioners of which 80% have coded professionally for at least 2 years, assess subjective perspectives based on experience. To validity, we compare situations in developers naturally different setups -- the difference between working home or office, and how things changed when were forced work from due Covid-19 pandemic. The results indicate that using...
In this paper, we present DeepSIM, a generative model for conditional image manipulation based on single image. We find that extensive augmentation is key enabling training, and incorporate the use of thin-plate-spline (TPS) as an effective augmentation. Our network learns to map between primitive representation itself. The choice has impact ease expressiveness manipulations can be automatic (e.g. edges), manual segmentation) or hybrid such edges top segmentations. At time, our generator...