Pretrained large-scale vision-language models such as CLIP have demonstrated excellent generalizability over a series of downstream tasks. However, they are sensitive to the variation of input text prompts and need a selection of prompt templates to achieve satisfactory performance. Recently, various methods have been proposed to dynamically learn the prompts as the textual inputs to avoid the requirements of laboring hand-crafted prompt engineering in the fine-tuning process. We notice that these methods are suboptimal in two aspects. First, the prompts of the vision and language branches in these methods are usually separated or uni-directionally correlated. Thus, the prompts of both branches are not fully correlated and may not provide enough guidance to align the representations of both branches. Second, it's observed that most previous methods usually achieve better performance on seen classes but cause performance degeneration on unseen classes compared to CLIP. This is because the essential generic knowledge learned in the pretraining stage is partly forgotten in the fine-tuning process. In this paper, we propose Co-Articulated Multi-Modal Learning (COMMA) to handle the above limitations. Especially, our method considers prompts from both branches to generate the prompts to enhance the representation alignment of both branches. Besides, to alleviate forgetting about the essential knowledge, we minimize the feature discrepancy between the learned prompts and the embeddings of hand-crafted prompts in the pre-trained CLIP in the late transformer layers. We evaluate our method across three representative tasks of generalization to novel classes, new target datasets and unseen domain shifts. Experimental results demonstrate the superiority of our method by exhibiting a favorable performance boost upon all tasks with high efficiency. Code is available at https://github.com/hulianyuyy/COMMA.

Skeleton-aware sign language recognition (SLR) has gained popularity due to its ability to remain unaffected by background information and its lower computational requirements. Current methods utilize spatial graph modules and temporal modules to capture spatial and temporal features, respectively. However, their spatial graph modules are typically built on fixed graph structures such as graph convolutional networks or a single learnable graph, which only partially explore joint relationships. Additionally, a simple temporal convolution kernel is used to capture temporal information, which may not fully capture the complex movement patterns of different signers. To overcome these limitations, we propose a new spatial architecture consisting of two concurrent branches, which build input-sensitive joint relationships and incorporates specific domain knowledge for recognition, respectively. These two branches are followed by an aggregation process to distinguishe important joint connections. We then propose a new temporal module to model multi-scale temporal information to capture complex human dynamics. Our method achieves state-of-the-art accuracy compared to previous skeleton-aware methods on four large-scale SLR benchmarks. Moreover, our method demonstrates superior accuracy compared to RGB-based methods in most cases while requiring much fewer computational resources, bringing better accuracy-computation trade-off. Code is available at https://github.com/hulianyuyy/DSTA-SLR.

Hand and face play an important role in expressing sign language. Their features are usually especially leveraged to improve system performance. However, to effectively extract visual representations and capture trajectories for hands and face, previous methods always come at high computations with increased training complexity. They usually employ extra heavy pose-estimation networks to locate human body keypoints or rely on additional pre-extracted heatmaps for supervision. To relieve this problem, we propose a self-emphasizing network (SEN) to emphasize informative spatial regions in a self-motivated way, with few extra computations and without additional expensive supervision. Specifically, SEN first employs a lightweight subnetwork to incorporate local spatial-temporal features to identify informative regions, and then dynamically augment original features via attention maps. It's also observed that not all frames contribute equally to recognition. We present a temporal self-emphasizing module to adaptively emphasize those discriminative frames and suppress redundant ones. A comprehensive comparison with previous methods equipped with hand and face features demonstrates the superiority of our method, even though they always require huge computations and rely on expensive extra supervision. Remarkably, with few extra computations, SEN achieves new state-of-the-art accuracy on four large-scale datasets, PHOENIX14, PHOENIX14-T, CSL-Daily, and CSL. Visualizations verify the effects of SEN on emphasizing informative spatial and temporal features. Code is available at https://github.com/hulianyuyy/SEN_CSLR

Human body trajectories are a salient cue to identify actions in the video. Such body trajectories are mainly conveyed by hands and face across consecutive frames in sign language. However, current methods in continuous sign language recognition (CSLR) usually process frames independently, thus failing to capture cross-frame trajectories to effectively identify a sign. To handle this limitation, we propose correlation network (CorrNet) to explicitly capture and leverage body trajectories across frames to identify signs. In specific, a correlation module is first proposed to dynamically compute correlation maps between the current frame and adjacent frames to identify trajectories of all spatial patches. An identification module is then presented to dynamically emphasize the body trajectories within these correlation maps. As a result, the generated features are able to gain an overview of local temporal movements to identify a sign. Thanks to its special attention on body trajectories, CorrNet achieves new state-of-the-art accuracy on four largescale datasets, i.e., PHOENIX14, PHOENIX14-T, CSL-Daily, and CSL. A comprehensive comparison with previous spatial-temporal reasoning methods verifies the effectiveness of CorrNet. Visualizations demonstrate the effects of CorrNet on emphasizing human body trajectories across adjacent frames.

Raw videos have been proven to own considerable feature redundancy where in many cases only a portion of frames can already meet the requirements for accurate recognition. In this paper, we are interested in whether such redundancy can be effectively leveraged to facilitate efficient inference in continuous sign language recognition (CSLR). We propose a novel adaptive model (AdaBrowse) to dynamically select a most informative subsequence from input video sequences by modelling this problem as a sequential decision task. In specific, we first utilize a lightweight network to quickly scan input videos to extract coarse features. Then these features are fed into a policy network to intelligently select a subsequence to process. The corresponding subsequence is finally inferred by a normal CSLR model for sentence prediction. As only a portion of frames are processed in this procedure, the total computations can be considerably saved. Besides temporal redundancy, we are also interested in whether the inherent spatial redundancy can be seamlessly integrated together to achieve further efficiency, i.e., dynamically selecting a lowest input resolution for each sample, whose model is referred to as AdaBrowse+. Extensive experimental results on four large-scale CSLR datasets, i.e., PHOENIX14, PHOENIX14-T, CSL-Daily and CSL, demonstrate the effectiveness of AdaBrowse and AdaBrowse+ by achieving comparable accuracy with state-of-the-art methods with 1.44X throughput and 2.12X fewer FLOPs. Comparisons with other commonly-used 2D CNNs and adaptive efficient methods verify the effectiveness of AdaBrowse. Code is available at https://github.com/hulianyuyy/AdaBrowse.