Generating image variations, where a model produces variations of an input image while preserving the semantic context has gained increasing attention. Current image variation techniques involve adapting a text-to-image model to reconstruct an input image conditioned on the same image. We first demonstrate that a diffusion model trained to reconstruct an input image from frozen embeddings, can reconstruct the image with minor variations. Second, inspired by how text-to-image models learn from web-scale text-image pairs, we explore a new pretraining strategy to generate image variations using a large collection of image pairs. Our diffusion model Semantica receives a random (encoded) image from a webpage as conditional input and denoises another noisy random image from the same webpage. We carefully examine various design choices for the image encoder, given its crucial role in extracting relevant context from the input image. Once trained, Semantica can adaptively generate new images from a dataset by simply using images from that dataset as input. Finally, we identify limitations in standard image consistency metrics for evaluating image variations and propose alternative metrics based on few-shot generation.

PaliGemma is an open Vision-Language Model (VLM) that is based on the SigLIP-So400m vision encoder and the Gemma-2B language model. It is trained to be a versatile and broadly knowledgeable base model that is effective to transfer. It achieves strong performance on a wide variety of open-world tasks. We evaluate PaliGemma on almost 40 diverse tasks including standard VLM benchmarks, but also more specialized tasks such as remote-sensing and segmentation.

Training a linear classifier or lightweight model on top of pretrained vision model outputs, so-called 'frozen features', leads to impressive performance on a number of downstream few-shot tasks. Currently, frozen features are not modified during training. On the other hand, when networks are trained directly on images, data augmentation is a standard recipe that improves performance with no substantial overhead. In this paper, we conduct an extensive pilot study on few-shot image classification that explores applying data augmentations in the frozen feature space, dubbed 'frozen feature augmentation (FroFA)', covering twenty augmentations in total. Our study demonstrates that adopting a deceptively simple pointwise FroFA, such as brightness, can improve few-shot performance consistently across three network architectures, three large pretraining datasets, and eight transfer datasets.

Perceptual distances between images, as measured in the space of pre-trained deep features, have outperformed prior low-level, pixel-based metrics on assessing perceptual similarity. While the capabilities of older and less accurate models such as AlexNet and VGG to capture perceptual similarity are well known, modern and more accurate models are less studied. In this paper, we present a large-scale empirical study to assess how well ImageNet classifiers perform on perceptual similarity. First, we observe a inverse correlation between ImageNet accuracy and Perceptual Scores of modern networks such as ResNets, EfficientNets, and Vision Transformers: that is better classifiers achieve worse Perceptual Scores. Then, we examine the ImageNet accuracy/Perceptual Score relationship on varying the depth, width, number of training steps, weight decay, label smoothing, and dropout. Higher accuracy improves Perceptual Score up to a certain point, but we uncover a Pareto frontier between accuracies and Perceptual Score in the mid-to-high accuracy regime. We explore this relationship further using a number of plausible hypotheses such as distortion invariance, spatial frequency sensitivity, and alternative perceptual functions. Interestingly we discover shallow ResNets and ResNets trained for less than 5 epochs only on ImageNet, whose emergent Perceptual Score matches the prior best networks trained directly on supervised human perceptual judgements. The checkpoints for the models in our study are available at https://console.cloud.google.com/storage/browser/gresearch/perceptual_similarity.

We propose Dual PatchNorm: two Layer Normalization layers (LayerNorms), before and after the patch embedding layer in Vision Transformers. We demonstrate that Dual PatchNorm outperforms the result of exhaustive search for alternative LayerNorm placement strategies in the Transformer block itself. In our experiments, incorporating this trivial modification, often leads to improved accuracy over well-tuned Vision Transformers and never hurts.

The present study was conducted to assess the expected genetic gain for first lactation production and reproduction traits in Murrah buffaloes, in addition to optimization of progenies/sire. Data for period 1971-2020 were used from National Dairy Research Institute. Performance traits considered were 305 days milk yield (305DMY), average daily milk yield (ADMY), peak yield (PY), lactation length (LL), calving to first insemination interval (CFI), days open (DO), and calving interval (CI). Expected ΔG was estimated and compared by three different methods; method I involved heritability and selection differential; method II involved selection intensity, phenotypic standard deviation, and heritability; method III involved estimation of ΔG through four paths of inheritance. Initially, eleven progenies/sire were utilized for assessing expected ΔG by method III, and expected ΔG was found as 34.33, 0.12, 0.12 kg, 2.63, 1.51, 2.74, and 2.80 days/year for 305DMY, ADMY, PY, LL, CFI, DO, and CI, respectively. Additionally, there was a significant increase in expected ΔG on increasing progenies/sire from 6 to 11 while subsequent increase upto 16 had little effect on expected ΔG. These findings will be helpful in formulating breeding strategies worldwide in small buffalo herds to obtain sustainable ΔG in production and reproduction traits.

Contrastive pretraining on image-text pairs from the web is one of the most popular large-scale pretraining strategies for vision backbones, especially in the context of large multimodal models. At the same time, image captioning on this type of data is commonly considered an inferior pretraining strategy. In this paper, we perform a fair comparison of these two pretraining strategies, carefully matching training data, compute, and model capacity. Using a standard encoder-decoder transformer, we find that captioning alone is surprisingly effective: on classification tasks, captioning produces vision encoders competitive with contrastively pretrained encoders, while surpassing them on vision language tasks. We further analyze the effect of the model architecture and scale, as well as the pretraining data on the representation quality, and find that captioning exhibits the same or better scaling behavior along these axes. Overall our results show that plain image captioning is a more powerful pretraining strategy than was previously believed.

. Perceptual distances between images, as measured in the space of pre-trained deep features, have outperformed prior low-level, pixel-based metrics on assessing image similarity. While the capabilities of older and less accurate models such as AlexNet and VGG to capture perceptual similarity are well known, modern and more accurate models are less studied. First, we observe a surprising inverse correlation between ImageNet accuracy and Perceptual Scores of modern networks such as ResNets, EfficientNets, and Vision Transformers: that is better classifiers achieve worse Perceptual Scores. Then, we perform a large-scale study and examine the ImageNet accuracy/Perceptual Score relationship on varying the depth, width, number of training steps, weight decay, label smoothing, and dropout. Higher accuracy improves Perceptual Score up to a certain point, but we uncover a Pareto frontier between accuracies and Perceptual Score in the mid-to-high accuracy regime. We explore this relationship further using distortion invariance, spatial frequency sensitivity, and alternative perceptual functions. Interestingly we discover shallow ResNets, trained for less than 5 epochs only on ImageNet, whose emergent Perceptual Score matches the prior best networks trained directly on supervised human perceptual judgements.

Existing weather forecasting models are based on physics and use supercomputers to evolve the atmosphere into the future. Better physics-based forecasts require improved atmospheric models, which can be difficult to discover and develop, or increasing the resolution underlying the simulation, which can be computationally prohibitive. An emerging class of weather models based on neural networks overcome these limitations by learning the required transformations from data instead of relying on hand-coded physics and by running efficiently in parallel. Here we present a neural network capable of predicting precipitation at a high resolution up to 12 h ahead. The model predicts raw precipitation targets and outperforms for up to 12 h of lead time state-of-the-art physics-based models currently operating in the Continental United States. The results represent a substantial step towards validating the new class of neural weather models.

The problem of forecasting weather has been scientifically studied for centuries due to its high impact on human lives, transportation, food production and energy management, among others. Current operational forecasting models are based on physics and use supercomputers to simulate the atmosphere to make forecasts hours and days in advance. Better physics-based forecasts require improvements in the models themselves, which can be a substantial scientific challenge, as well as improvements in the underlying resolution, which can be computationally prohibitive. An emerging class of weather models based on neural networks represents a paradigm shift in weather forecasting: the models learn the required transformations from data instead of relying on hand-coded physics and are computationally efficient. For neural models, however, each additional hour of lead time poses a substantial challenge as it requires capturing ever larger spatial contexts and increases the uncertainty of the prediction. In this work, we present a neural network that is capable of large-scale precipitation forecasting up to twelve hours ahead and, starting from the same atmospheric state, the model achieves greater skill than the state-of-the-art physics-based models HRRR and HREF that currently operate in the Continental United States. Interpretability analyses reinforce the observation that the model learns to emulate advanced physics principles. These results represent a substantial step towards establishing a new paradigm of efficient forecasting with neural networks.

We present the Colorization Transformer, a novel approach for diverse high fidelity image colorization based on self-attention. Given a grayscale image, the colorization proceeds in three steps. We first use a conditional autoregressive transformer to produce a low resolution coarse coloring of the grayscale image. Our architecture adopts conditional transformer layers to effectively condition grayscale input. Two subsequent fully parallel networks upsample the coarse colored low resolution image into a finely colored high resolution image. Sampling from the Colorization Transformer produces diverse colorings whose fidelity outperforms the previous state-of-the-art on colorising ImageNet based on FID results and based on a human evaluation in a Mechanical Turk test. Remarkably, in more than 60% of cases human evaluators prefer the highest rated among three generated colorings over the ground truth. The code and pre-trained checkpoints for Colorization Transformer are publicly available at https://github.com/google-research/google-research/tree/master/coltran

Generative models that can model and predict sequences of future events can, in principle, learn to capture complex real-world phenomena, such as physical interactions. However, a central challenge in video prediction is that the future is highly uncertain: a sequence of past observations of events can imply many possible futures. Although a number of recent works have studied probabilistic models that can represent uncertain futures, such models are either extremely expensive computationally as in the case of pixel-level autoregressive models, or do not directly optimize the likelihood of the data. To our knowledge, our work is the first to propose multi-frame video prediction with normalizing flows, which allows for direct optimization of the data likelihood, and produces high-quality stochastic predictions. We describe an approach for modeling the latent space dynamics, and demonstrate that flow-based generative models offer a viable and competitive approach to generative modelling of video.

Data on first lactation reproduction traits of Sahiwal cattle maintained at an organized herd of the institute over a period of 29 years (1988–2016) were analyzed in the present study. Effect of various genetic and non–genetic factors on first lactation reproduction traits in addition to their genetic control was assessed by using mixed model least–square analysis. The overall least–square means for first lactation reproduction traits, viz. age at first service (AFS), age at first calving (AFC), number of services per conception (NSC), conception rate (CR), gestation period (GP), calving to first insemination interval (CFI), service period (SP), calving interval (CI) and pregnancy rate (PR) were 784.84±8.03 days, 1116.18±10.38 days, 1.86±0.11, 0.97±0.06, 287.73±0.60, 84.75±5.01, 119.86±8.31, 420.53±3.33 days and 0.27±0.02, respectively. Analysis of variance revealed that random effect of sire had non– significant influence on all the considered traits in present study. Period of birth significantly influenced AFS and AFC. Differences were statistically significant for effect of period of calving on NSC; season of calving on NSC, CR and PR; age group on NSC and CR. The effect of lactation length used as co–variable had significant influence on CFI, SP and PR. Service period as co–variable significantly affected CI. The magnitudes of heritability estimates were low to medium varying from 0.09 to 0.27 for different reproduction traits indicating large role of environment and management for improvement of these traits.

In the present review, an attempt has been made to explore Murrah buffalo, an important milch breed for its performance, genetic parameters, constraints and future scenario. In India, price of milk is decided on basis of fat percentage and buffalo milk fetches more money compared to cow milk to dairy farmers. Production and reproduction performances are among most important considerations to determine the profitability of any dairy farm. The production, reproduction and functional traits of Murrah buffaloes as reported by different workers at different organized farm were reviewed. Literature revealed that the 305-days lactation milk yield, 305-days lactation fat yield, 305-days lactation SNF yield, 305-days lactation total solid yield, peak yield, Fat %, SNF %, lactation length, dry period, age at first calving, service period, conception rate, pregnancy rate and calving interval varied from 1365±03 to 2086.17±44.66 kg, 118.3 to 167.38±4.46 kg,185.5±3.4 to 198.88±5.05 kg, 336.2±6.1 to 360.61±9.25 kg, 9.96 to 11.13±0.44 kg, 6.84% to 8.17%, 9.36% to 9.76%, 267.15 to 321.62 days,121.68 to 250.5 days, 1202 to 1618.83 days, 135.79 to 308 days, 68.80%, and 428.30 to 559.6 days, respectively in Murrah buffaloes. Heritability and repeatability estimates for the production, 38% reproduction and functional traits of Murrah buffaloes were also reviewed. Breeding, feeding, health and housing management practices have much impact on production and reproduction performance of animals and ultimately influence the economy of dairy farmers. Constraints in all these aspects are the obstacles to implement better animal husbandry practices in dairy animals and these should be overcome by taking suitable measures.

We present a simple and powerful algorithm for parallel black box optimization called Successive Halving and Classification (SHAC). The algorithm operates in $K$ stages of parallel function evaluations and trains a cascade of binary classifiers to iteratively cull the undesirable regions of the search space. SHAC is easy to implement, requires no tuning of its own configuration parameters, is invariant to the scale of the objective function and can be built using any choice of binary classifier. We adopt tree-based classifiers within SHAC and achieve competitive performance against several strong baselines for optimizing synthetic functions, hyperparameters and architectures.

Present study was carried out with the aim to determine the effect of genetic and non-genetic factors on milk production efficiency traits and their genetic control. Data on milk production efficiency traits i.e. total milk yield per day of lactation length and total milk yield per day of calving interval in Jersey crossbred animals maintained at organized herd of ICARNational Dairy Research Institute, Eastern Regional Station, Kalyani, Nadia, West Bengal spread over 29 years (1986-2014) were analyzed in the study. The overall least-squares means of total milk yield per day of lactation length and total milk yield per day of calving interval were estimated as 6.97±0.21 and 6.02±0.23 kg/day, respectively. The data was grouped into different sub-classes of season of calving, period of calving, parity and genetic group of animals. The random effect of sire had shown significant (p<0.01) influence on milk production efficiency traits. Various environment factors significantly affected the milk production efficiency traits indicating the scope of improvement in management of these crossbred animals. The heritability estimates of 0.54 for TMY/LL and 0.63 for TMY/CI were high in magnitude which suggests that fast genetic progress can be made, if these animals are selected based on the milk production efficiency traits.

PCR amplification of exon-2 of leptin gene yielded an amplified product of 210 bp in Murrah buffaloes. RFLP of leptin gene done with Aci1 restriction enzyme revealed 3 types of genotypes CC, TC and TT in Murrah buffaloes of NDRI herd with a frequency of 0.26 , 0.73 and 0.013% respectively. Results suggested that CC genotype of leptin gene in Murrah buffaloes is significantly associated with 305 days milk yield. Overall study suggested the scope of considering genetic information of leptin gene in breeding and management strategies for the improvement of Murrah buffaloes of NDRI herd.

Aim:The present study was performed to investigate the effect of genetic and non-genetic factors affecting milk fat yield and to estimate genetic parameters of monthly test day fat yields (MTDFY) and lactation 305-day fat yield (L305FY) in Murrah buffaloes. Materials and Methods:The data on total of 10381 MTDFY records comprising the first four lactations of 470 Murrah buffaloes calved from 1993 to 2014 were assessed.These buffaloes were sired by 75 bulls maintained in an organized farm at ICAR-National Dairy Research Institute, Karnal.Least squares maximum likelihood program was used to estimate genetic and non-genetic parameters.Heritability estimates were obtained using paternal half-sib correlation method.Genetic and phenotypic correlations among MTDFY, and 305-day fat yield were calculated from the analysis of variance and covariance matrix among sire groups. Results:The overall least squares mean of L305FY was found to be 175.74±4.12kg.The least squares mean of overall MTDFY ranged from 3.33±0.14kg (TD-11) to 7.06±0.17kg (TD-3).The h 2 estimate of L305FY was found to be 0.33±0.16 in this study.The estimates of phenotypic and genetic correlations between 305-day fat yield and different MTDFY ranged from 0.32 to 0.48 and 0.51 to 0.99, respectively. Conclusions:In this study, all the genetic and non-genetic factors except age at the first calving group, significantly affected the traits under study.The estimates of phenotypic and genetic correlations of MTDFY with 305-day fat yield was generally higher in the MTDFY-5 of lactation suggesting that this TD yields could be used as the selection criteria for early evaluation and selection of Murrah buffaloes.