Search for: All records

Since American Sign Language (ASL) has no standard written form, Deaf signers frequently share videos in order to communicate in their native language. However, this does not preserve privacy. Since critical linguistic information is transmitted through facial expressions, the face cannot be obscured. While signers have expressed interest, for a variety of applications, in sign language video anonymization that would effectively preserve linguistic content, attempts to develop such technology have had limited success and generally require pose estimation that cannot be readily carried out in the wild. To address current limitations, our research introduces DiffSLVA, a novel methodology that uses pre-trained large-scale diffusion models for text-guided sign language video anonymization. We incorporate ControlNet, which leverages low-level image features such as HED (Holistically-Nested Edge Detection) edges, to circumvent the need for pose estimation. Additionally, we develop a specialized module to capture linguistically essential facial expressions. We then combine the above methods to achieve anonymization that preserves the essential linguistic content of the original signer. This innovative methodology makes possible, for the first time, sign language video anonymization that could be used for real-world applications, which would offer significant benefits to the Deaf and Hard-of-Hearing communities. 

We propose a multimodal network using skeletons and handshapes as input to recognize individual signs and detect their boundaries in American Sign Language (ASL) videos. Our method integrates a spatio-temporal Graph Convolutional Network (GCN) architecture to estimate human skeleton keypoints; it uses a late-fusion approach for both forward and backward processing of video streams. Our (core) method is designed for the extraction---and analysis of features from---ASL videos, to enhance accuracy and efficiency of recognition of individual signs. A Gating module based on per-channel multi-layer convolutions is employed to evaluate significant frames for recognition of isolated signs. Additionally, an auxiliary multimodal branch network, integrated with a transformer, is designed to estimate the linguistic start and end frames of an isolated sign within a video clip. We evaluated performance of our approach on multiple datasets that include isolated, citation-form signs and signs pre-segmented from continuous signing based on linguistic annotations of start and end points of signs within sentences. We have achieved very promising results when using both types of sign videos combined for training, with overall sign recognition accuracy of 80.8% Top-1 and 95.2% Top-5 for citation-form signs, and 80.4% Top-1 and 93.0% Top-5 for signs pre-segmented from continuous signing. 

Since American Sign Language (ASL) has no standard written form, Deaf signers frequently share videos in order to communicate in their native language. However, since both hands and face convey critical linguistic information in signed languages, sign language videos cannot preserve signer privacy. While signers have expressed interest, for a variety of applications, in sign language video anonymization that would effectively preserve linguistic content, attempts to develop such technology have had limited success, given the complexity of hand movements and facial expressions. Existing approaches rely predominantly on precise pose estimations of the signer in video footage and often require sign language video datasets for training. These requirements prevent them from processing videos 'in the wild,' in part because of the limited diversity present in current sign language video datasets. To address these limitations, our research introduces DiffSLVA, a novel methodology that utilizes pre-trained large-scale diffusion models for zero-shot text-guided sign language video anonymization. We incorporate ControlNet, which leverages low-level image features such as HED (Holistically-Nested Edge Detection) edges, to circumvent the need for pose estimation. Additionally, we develop a specialized module dedicated to capturing facial expressions, which are critical for conveying essential linguistic information in signed languages. We then combine the above methods to achieve anonymization that better preserves the essential linguistic content of the original signer. This innovative methodology makes possible, for the first time, sign language video anonymization that could be used for real-world applications, which would offer significant benefits to the Deaf and Hard-of-Hearing communities. We demonstrate the effectiveness of our approach with a series of signer anonymization experiments. 

The adversarial risk of a machine learning model has been widely studied. Most previous studies assume that the data lie in the whole ambient space. We propose to take a new angle and take the manifold assumption into consideration. Assuming data lie in a manifold, we investigate two new types of adversarial risk, the normal adversarial risk due to perturbation along normal direction and the in-manifold adversarial risk due to perturbation within the manifold. We prove that the classic adversarial risk can be bounded from both sides using the normal and in-manifold adversarial risks. We also show a surprisingly pessimistic case that the standard adversarial risk can be non-zero even when both normal and in-manifold adversarial risks are zero. We finalize the study with empirical studies supporting our theoretical results. Our results suggest the possibility of improving the robustness of a classifier without sacrificing model accuracy, by only focusing on the normal adversarial risk. 

The adversarial risk of a machine learning model has been widely studied. Most previous works assume that the data lies in the whole ambient space. We propose to take a new angle and take the manifold assumption into consideration. Assuming data lies in a manifold, we investigate two new types of adversarial risk, the normal adversarial risk due to perturbation along normal direction, and the in-manifold adversarial risk due to perturbation within the manifold. We prove that the classic adversarial risk can be bounded from both sides using the normal and in-manifold adversarial risks. We also show with a surprisingly pessimistic case that the standard adversarial risk can be nonzero even when both normal and in-manifold risks are zero. We finalize the paper with empirical studies supporting our theoretical results. Our results suggest the possibility of improving the robustness of a classifier by only focusing on the normal adversarial risk. 

To improve computer-based recognition from video of isolated signs from American Sign Language (ASL), we propose a new skeleton-based method that involves explicit detection of the start and end frames of signs, trained on the ASLLVD dataset; it uses linguistically relevant parameters based on the skeleton input. Our method employs a bidirectional learning approach within a Graph Convolutional Network (GCN) framework. We apply this method to the WLASL dataset, but with corrections to the gloss labeling to ensure consistency in the labels assigned to different signs; it is important to have a 1-1 correspondence between signs and text-based gloss labels. We achieve a success rate of 77.43% for top-1 and 94.54% for top-5 using this modified WLASL dataset. Our method, which does not require multi-modal data input, outperforms other state-of-the-art approaches on the same modified WLASL dataset, demonstrating the importance of both attention to the start and end frames of signs and the use of bidirectional data streams in the GCNs for isolated sign recognition. 

We present a new approach for isolated sign recognition, which combines a spatial-temporal Graph Convolution Network (GCN) architecture for modeling human skeleton keypoints with late fusion of both the forward and backward video streams, and we explore the use of curriculum learning. We employ a type of curriculum learning that dynamically estimates, during training, the order of difficulty of each input video for sign recognition; this involves learning a new family of data parameters that are dynamically updated during training. The research makes use of a large combined video dataset for American Sign Language (ASL), including data from both the American Sign Language Lexicon Video Dataset (ASLLVD) and the Word-Level American Sign Language (WLASL) dataset, with modified gloss labeling of the latter—to ensure 1-1 correspondence between gloss labels and distinct sign productions, as well as consistency in gloss labeling across the two datasets. This is the first time that these two datasets have been used in combination for isolated sign recognition research. We also compare the sign recognition performance on several different subsets of the combined dataset, varying in, e.g., the minimum number of samples per sign (and therefore also in the total number of sign classes and video examples). 

We present a new approach for isolated sign recognition, which combines a spatial-temporal Graph Convolution Network (GCN) architecture for modeling human skeleton keypoints with late fusion of both the forward and backward video streams, and we explore the use of curriculum learning. We employ a type of curriculum learning that dynamically estimates, during training, the order of difficulty of each input video for sign recognition; this involves learning a new family of data parameters that are dynamically updated during training. The research makes use of a large combined video dataset for American Sign Language (ASL), including data from both the American Sign Language Lexicon Video Dataset (ASLLVD) and the Word-Level American Sign Language (WLASL) dataset, with modified gloss labeling of the latter—to ensure 1-1 correspondence between gloss labels and distinct sign productions, as well as consistency in gloss labeling across the two datasets. This is the first time that these two datasets have been used in combination for isolated sign recognition research. We also compare the sign recognition performance on several different subsets of the combined dataset, varying in, e.g., the minimum number of samples per sign (and therefore also in the total number of sign classes and video examples). 

Deaf signers who wish to communicate in their native language frequently share videos on the Web. However, videos cannot preserve privacy—as is often desirable for discussion of sensitive topics—since both hands and face convey critical linguistic information and therefore cannot be obscured without degrading communication. Deaf signers have expressed interest in video anonymization that would preserve linguistic content. However, attempts to develop such technology have thus far shown limited success. We are developing a new method for such anonymization, with input from ASL signers. We modify a motion-based image animation model to generate high-resolution videos with the signer identity changed, but with preservation of linguistically significant motions and facial expressions. An asymmetric encoder-decoder structured image generator is used to generate the high-resolution target frame from the low-resolution source frame based on the optical flow and confidence map. We explicitly guide the model to attain clear generation of hands and face by using bounding boxes to improve the loss computation. FID and KID scores are used for evaluation of the realism of the generated frames. This technology shows great potential for practical applications to benefit deaf signers. 

Deaf signers who wish to communicate in their native language frequently share videos on the Web. However, videos cannot preserve privacy—as is often desirable for discussion of sensitive topics—since both hands and face convey critical linguistic information and therefore cannot be obscured without degrading communication. Deaf signers have expressed interest in video anonymization that would preserve linguistic content. However, attempts to develop such technology have thus far shown limited success. We are developing a new method for such anonymization, with input from ASL signers. We modify a motion-based image animation model to generate high-resolution videos with the signer identity changed, but with preservation of linguistically significant motions and facial expressions. An asymmetric encoder-decoder structured image generator is used to generate the high-resolution target frame from the low-resolution source frame based on the optical flow and confidence map. We explicitly guide the model to attain clear generation of hands and face by using bounding boxes to improve the loss computation. FID and KID scores are used for evaluation of the realism of the generated frames. This technology shows great potential for practical applications to benefit deaf signers.