ACML 2024 Papers

The mainstream workflow of image recognition applications is first training one global model on the cloud for a wide range of classes and then serving numerous clients. Images uploaded by each client typically come from a small subset of classes. From the cloud-client discrepancy on the range of image classes, the recognition model is desired to have strong adaptiveness, intuitively by focusing on each client’s local dynamic class subset, while incurring negligible overhead. In this work, we propose to plug a new intra-client and inter-image attention (ICIIA) module into existing backbone recognition models, requiring only one-time cloud-based training to be client-adaptive. In particular, given an image to be recognized from a certain client, ICIIA introduces multi-head self-attention to retrieve relevant images from the client’s local images, thereby calibrating the focus and the recognition result. We further identify the bottleneck of ICIIA’s overhead being in linear projection, propose to group and shuffle the features before the projections, and allow increasing the number of feature groups to dramatically improve efficiency without scarifying much accuracy. We extensively evaluate ICIIA and compare its performance against several baselines, demonstrating effectiveness and efficiency. Specifically, for a partitioned version of ImageNet-1K with the backbone models of MobileNetV3-L and Swin-B, ICIIA improves the classification accuracy to 83.37% (+8.11%) and 88.86% (+5.28%), while adding only 1.62% and 0.02% of FLOPs, respectively. Source code is available in the supplementary materials.

Audio-Visual Speech Recognition (AVSR) is a task that identifies spoken words by analyzing both lip movements and auditory signals. Compared to Automatic Speech Recognition (ASR), AVSR demonstrates greater robustness in noisy environments due to the support of dual modalities. However, the inherent differences between these modalities present a challenge: effectively accounting for their disparities and leveraging their complementary information to extract useful information for AVSR. To address this, we propose the AMG-AVSR model, which utilizes a two-stage curriculum learning strategy and incorporates a feature compression and recovery mechanism. By leveraging the characteristics of different modalities in various scenarios to guide each other, the model extracts refined features from audio-visual data, thereby enhancing recognition performance in both clean and noisy environments. Compared to the baseline model AV-HuBERT, AMG-AVSR demonstrates superior performance on the LRS2 dataset in both noisy and clean environments. AMG-AVSR achieves a word error rate (WER) of 2.9% under clean speech conditions. In various noisy conditions, AMG-AVSR shows a significant reduction in WER compared to previous methods.

Automatic license plate recognition is applied widely in life, but it is still a challenging task in open scenarios. The current ALPR methods require multiple recognitions in multi-license plate scenarios. Furthermore, they have complex recognition structures and insufficient recognition capabilities for multi-style license plates. To solve the above problems, this paper proposes a single-stage multi-style multi-license plate recognition method based on the attention mechanism: CLPRNet. We use a spatial attention module based on UNet to separate the license plate character sequence into each attention heatmap in order. This approach unifies license plates with different character lengths and different character rows into a single processing logic, thereby enabling CLPRNet to recognize multiple styles without additional style judgement branches. At the same time, we abandon the traditional method of cropping RoI from image or feature, and instead combine attention to recognize characters directly, which allows CLPRNet to recognize multiple license plates in a single pass. To address the issue of an inadequate number of multi-style license plate samples, this paper also proposes a multi-style license plate generation method. In the single-stage methods, CLPRNet demonstrates better detection performance on the CCPD dataset and better recognition performance on the FN, Rotate, and Tilt subsets of the CCPD dataset. Compared with the existing license plate recognition methods, CLPRNet can recognize more styles of license plates. Test results and ablation experiments have shown the effectiveness of our proposed method.

Similar case retrieval is a crucial aspect of the legal retrieval field, significantly contributing to LegalAI systems. This task aims to retrieve cases that are highly relevant to the query case, thereby enhancing the efficiency of legal practitioners. Recent methods have leveraged the rich semantic knowledge of pre-trained models, greatly improving retrieval performance. However, these methods often overlook key legal elements within the complex language structures of case texts, such as legal event information that can impact case outcomes and judgments. This oversight results in the underutilization of critical case information. To address this issue, we proposed RAEvent, a similar case retrieval contrastive framework augmented by legal event information. This framework utilizes an enhanced case event information database to provide auxiliary information for case retrieval and employs contrastive learning techniques to better extract similar features in cases. In comparison to a range of baseline approaches, the results of our experiments highlight the efficacy of our framework. Moreover, our research provides fresh perspectives and makes a valuable contribution to the ongoing studies in similar case retrieval tasks.

Microarray-generated genomic data has recently sparked a wave of bioinformatics and data mining research. However, such data presents significant challenges for further analysis due to its high dimensionality and small sample sizes. Feature selection is a standard approach to address this issue, as it can enhance classification performance while reducing dimensionality. This paper introduces an Improved Gray Wolf Optimization-based Evolutionary Multitasking (EMT-IGWO) feature selection approach tailored for high-dimensional classification. It adopts multi-population co-evolving searching modes that can be regarded as a typical feature selection task via a specific information-sharing mechanism. Within the proposed multitasking framework, both population diversity and global searching capabilities of EMT-IGWO are improved. Moreover, several enhancements are incorporated into the two searching modes to help stagnant individuals escape from local optima with higher probabilities. Computational results show that EMT-IGWO outperforms other compared algorithms in effectiveness and efficiency evaluated across eight public gene expression datasets.

With the widespread adoption of the Internet and mobile Internet, an increasing number of individuals are expressing their emotions on short-video platforms. Contemporary multimodal emotion analysis technologies facilitate a more comprehensive recognition and understanding of emotions through the analysis of various data sources including text, facial expressions, audio, hand gestures, among others. Consequently, the significance of sentiment analysis is becoming increasingly pronounced. However, existing research indicates that most emotion analysis techniques are not sufficiently rapid and efficient in light of the exponential proliferation of short video content. In addition, most sentiment analysis models demonstrate significant differences in the contribution of each modality, with text and visual modalities often exerting a greater influence than audio modes. Furthermore, in the pursuit of heightened accuracy, certain models are designed to be exceedingly complex, while others prioritize swift reasoning at the expense of accuracy. This paper proposes a more efficient multimodal sentiment analysis model, presenting three distinct advantages. Firstly, residual-free connectivity modules capable of extracting 3-D attentional weights are proposed to process visual modal features, maintaining accuracy while improving inference efficiency. Secondly, adoption of multi-scale hierarchical context aggregation (aggregation followed by interaction) for audio modality to capture coarse- and fine-grained audio contextual information through multilevel aggregation, thereby enriching audio modality features and minimizing disparities between modalities’ contributions. Finally, attainment of a superior balance between accuracy and speed, thereby enhancing adaptability to the fast-paced short video environment and meeting the burgeoning demand for video content processing.