Current benchmarks for sign language recognition (SLR) focus mainly on isolated SLR, while there are limited datasets for continuous SLR (CSLR), which recognizes sequences of signs in a video. Additionally, existing CSLR datasets are collected in controlled settings, which restricts their effectiveness in building robust real-world CSLR systems. To address these limitations, we present Isharah, a large multi-scene dataset for CSLR. It is the first dataset of its type and size, collected in an unconstrained environment using signers' smartphones. This setup resulted in high variations of recording settings, camera distances, angles, and resolutions. This variation helps with developing sign language understanding models capable of handling the variability and complexity of real-world scenarios. The dataset consists of 30,000 video clips performed by 18 deaf and professional signers. Additionally, the dataset is linguistically rich as it provides a gloss-level annotation for all dataset's videos, making it useful for developing CSLR and sign language translation (SLT) systems. This paper also introduces multiple sign language understanding benchmarks, including signer-independent and unseen-sentence CSLR, along with gloss-based and gloss-free SLT.

We explore different design choices for injecting noise into generative adversarial networks (GANs) with the goal of disentangling the latent space. Instead of traditional approaches, we propose feeding multiple noise codes through separate fully-connected layers respectively. The aim is restricting the influence of each noise code to specific parts of the generated image. We show that disentanglement in the first layer of the generator network leads to disentanglement in the generated image. Through a grid-based structure, we achieve several aspects of disentanglement without complicating the network architecture and without requiring labels. We achieve spatial disentanglement, scale-space disentanglement, and disentanglement of the foreground object from the background style allowing fine-grained control over the generated images. Examples include changing facial expressions in face images, changing beak length in bird images, and changing car dimensions in car images. This empirically leads to better disentanglement scores than state-of-the-art methods on the FFHQ dataset.

In multimodal unsupervised image-to-image translation tasks, the goal is to translate an image from the source domain to many images in the target domain. We present a simple method that produces higher quality images than current state-of-the-art while maintaining the same amount of multimodal diversity. Previous methods follow the unconditional approach of trying to map the latent code directly to a full-size image. This leads to complicated network architectures with several introduced hyperparameters to tune. By treating the latent code as a modifier of the convolutional filters, we produce multimodal output while maintaining the traditional Generative Adversarial Network (GAN) loss and without additional hyperparameters. The only tuning required by our method controls the tradeoff between variability and quality of generated images. Furthermore, we achieve disentanglement between source domain content and target domain style for free as a by-product of our formulation. We perform qualitative and quantitative experiments showing the advantages of our method compared with the state-of-the art on multiple benchmark image-to-image translation datasets.