Since the first Spanish settlers brought horses to America centuries ago, several local varieties and breeds have been established in the New World. These were generally a consequence of the admixture of the different breeds arriving from Europe. In some instances, local horses have been selectively bred for specific traits, such as appearance, endurance, strength, and gait. We looked at the genetics of two breeds, the Puerto Rican Non-Purebred (PRNPB) (also known as the “Criollo”) horses and the Puerto Rican Paso Fino (PRPF), from the Caribbean Island of Puerto Rico. While it is reasonable to assume that there was a historic connection between the two, the genetic link between them has never been established. In our study, we started by looking at the genetic ancestry and diversity of current Puerto Rican horse populations using a 668 bp fragment of the mitochondrial DNA D-loop (HVR1) in 200 horses from 27 locations on the island. We then genotyped all 200 horses in our sample for the “gait-keeper”
Equine Pain Behavior Classification via Self-Supervised Disentangled Pose Representation
Timely detection of horse pain is important for equine welfare. Horses express pain through their facial and body behavior, but may hide signs of pain from unfamiliar human observers. In addition, collecting visual data with detailed annotation of horse behavior and pain state is both cumbersome and not scalable. Consequently, a pragmatic equine pain classification system would use video of the unobserved horse and weak labels. This paper proposes such a method for equine pain classification by using multi-view surveillance video footage of unobserved horses with induced orthopaedic pain, with temporally sparse video level pain labels. To ensure that pain is learned from horse body language alone, we first train a self-supervised generative model to disentangle horse pose from its appearance and background before using the disentangled horse pose latent representation for pain classification. To make best use of the pain labels, we develop a novel loss that formulates pain classification as a multi-instance learning problem. Our method achieves pain classification accuracy better than human expert performance with 60% accuracy. The learned latent horse pose representation is shown to be viewpoint covariant, and disentangled from horse appearance. Qualitative analysis of pain classified segments shows correspondence between the pain symptoms identified by our model, and equine pain scales used in veterinary practice.
more »
« less
- Award ID(s):
- 2204808
- NSF-PAR ID:
- 10385711
- Date Published:
- Journal Name:
- Winter Conference on Applications of Computer Vision (WACV)
- Page Range / eLocation ID:
- 152 to 162
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
more » « less
Abstract DMRT3 mutant allele previously associated with the paso gait especially cherished in this island breed. We also genotyped a subset of 24 samples with the Illumina Neogen Equine Community genome-wide array (65,000 SNPs). This data was further combined with the publicly available PRPF genomes from other studies. Our analysis show an undeniable genetic connection between the two varieties in Puerto Rico, consistent with the hypothesis that PRNPB horses represent the descendants of the original genetic pool, a mix of horses imported from the Iberian Peninsula and elsewhere in Europe. Some of the original founders of PRNRB population must have carried the “gait-keeper”DMRT3 allele upon arrival to the island. From this admixture, the desired traits were selected by the local people over the span of centuries. We propose that the frequency of the mutant “gait-keeper” allele originally increased in the local horses due to the selection for the smooth ride and other characters, long before the PRPF breed was established. To support this hypothesis, we demonstrate that PRNPB horses, and not the purebred PRPF, carry a signature of selection in the genomic region containing theDMRT3 locus to this day. The lack of the detectable signature of selection associated with theDMRT3 in the PRPF would be expected if this native breed was originally derived from the genetic pool of PRNPB horses established earlier and most of the founders already had the mutant allele. Consequently, selection specific to PRPF later focused on allels in other genes (includingCHRM5, CYP2E1, MYH7, SRSF1, PAM, PRN and others) that have not been previously associated with the prized paso gait phenotype in Puerto Rico or anywhere else. -
Vedaldi, Andrea ; Bischof, Horst ; Brox, Thomas ; Frahm, Jan-Michael (Ed.)Novel view video synthesis aims to synthesize novel viewpoints videos given input captures of a human performance taken from multiple reference viewpoints and over consecutive time steps. Despite great advances in model-free novel view synthesis, existing methods present three limitations when applied to complex and time-varying human performance. First, these methods (and related datasets) mainly consider simple and symmetric objects. Second, they do not enforce explicit consistency across generated views. Third, they focus on static and non-moving objects. The fine-grained details of a human subject can therefore suffer from inconsistencies when synthesized across different viewpoints or time steps. To tackle these challenges, we introduce a human-specific framework that employs a learned 3D-aware representation. Specifically, we first introduce a novel siamese network that employs a gating layer for better reconstruction of the latent volumetric representation and, consequently, final visual results. Moreover, features from consecutive time steps are shared inside the network to improve temporal consistency. Second, we introduce a novel loss to explicitly enforce consistency across generated views both in space and in time. Third, we present the Multi-View Human Action (MVHA) dataset, consisting of near 1200 synthetic human performance captured from 54 viewpoints. Experiments on the MVHA, Pose-Varying Human Model and ShapeNet datasets show that our method outperforms the state-of-the-art baselines both in view generation quality and spatio-temporal consistency.more » « less
-
Logical properties such as negation, implication, and symmetry, despite the fact that they are foundational and threaded through the vocabulary and syntax of known natural languages, pose a special problem for language learning. Their meanings are much harder to identify and isolate in the child’s everyday interaction with referents in the world than concrete things (like spoons and horses) and happenings and acts (like running and jumping) that are much more easily identified, and thus more easily linked to their linguistic labels (spoon, horse, run, jump). Here we concentrate attention on the category of symmetry [a relation R is symmetrical if and only if (iff) for all x, y: if R ( x, y), then R (y, x)], expressed in English by such terms as similar, marry, cousin, and near. After a brief introduction to how symmetry is expressed in English and other well-studied languages, we discuss the appearance and maturation of this category in Nicaraguan Sign Language (NSL). NSL is an emerging language used as the primary, daily means of communication among a population of deaf individuals who could not acquire the surrounding spoken language because they could not hear it, and who were not exposed to a preexisting sign language because there was none available in their community. Remarkably, these individuals treat symmetry, in both semantic and syntactic regards, much as do learners exposed to a previously established language. These findings point to deep human biases in the structures underpinning and constituting human language.more » « less
-
Despite the advances in Human Activity Recognition, the ability to exploit the dynamics of human body motion in videos has yet to be achieved. In numerous recent works, re- searchers have used appearance and motion as independent inputs to infer the action that is taking place in a specific video. In this paper, we highlight that while using a novel representation of human body motion, we can benefit from appearance and motion simultaneously. As a result, bet- ter performance of action recognition can be achieved. We start with a pose estimator to extract the location and heat- map of body joints in each frame. We use a dynamic encoder to generate a fixed size representation from these body joint heat-maps. Our experimental results show that training a convolutional neural network with the dynamic motion representation outperforms state-of-the-art action recognition models. By modeling distinguishable activities as distinct dynamical systems and with the help of two stream net- works, we obtain the best performance on HMDB, JHMDB, UCF-101, and AVA datasets.more » « less
-
null (Ed.)Disentangled generative models map a latent code vector to a target space, while enforcing that a subset of the learned latent codes are interpretable and associated with distinct properties of the target distribution. Recent advances have been dominated by Variational AutoEncoder (VAE)-based methods, while training disentangled generative adversarial networks (GANs) remains challenging. In this work, we show that the dominant challenges facing disentangled GANs can be mitigated through the use of self-supervision. We make two main contributions: first, we design a novel approach for training disentangled GANs with self-supervision. We propose contrastive regularizer, which is inspired by a natural notion of disentanglement: latent traversal. This achieves higher disentanglement scores than state-of-the-art VAE- and GAN-based approaches. Second, we propose an unsupervised model selection scheme called ModelCentrality, which uses generated synthetic samples to compute the medoid (multi-dimensional generalization of median) of a collection of models. The current common practice of hyper-parameter tuning requires using ground-truths samples, each labelled with known perfect disentangled latent codes. As real datasets are not equipped with such labels, we propose an unsupervised model selection scheme and show that it finds a model close to the best one, for both VAEs and GANs. Combining contrastive regularization with ModelCentrality, we improve upon the state-of-the-art disentanglement scores significantly, without accessing the supervised data.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- https://doi.org/10.1109/wacv51458.2022.00023
