MaskMamba: A Hybrid Mamba-Transformer Model for Masked Image Generation

Image generation models have encountered challenges related to scalability and quadratic complexity, primarily due to the reliance on Transformer-based backbones. In this study, we introduce MaskMamba, a novel hybrid model that combines Mamba and Transformer architectures, utilizing Masked Image Modeling for non-autoregressive image synthesis. We meticulously redesign the bidirectional Mamba architecture by implementing two key modifications: (1) replacing causal convolutions with standard convolutions to better capture global context, and (2) utilizing concatenation instead of multiplication, which significantly boosts performance while accelerating inference speed. Additionally, we explore various hybrid schemes of MaskMamba, including both serial and grouped parallel arrangements. Furthermore, we incorporate an in-context condition that allows our model to perform both class-to-image and text-to-image generation tasks. Our MaskMamba outperforms Mamba-based and Transformer-based models in generation quality. Notably, it achieves a remarkable $54.44\%$ improvement in inference speed at a resolution of $2048\times 2048$ over Transformer.

Translatotron-V(ison): An End-to-End Model for In-Image Machine Translation

In-image machine translation (IIMT) aims to translate an image containing texts in source language into an image containing translations in target language. In this regard, conventional cascaded methods suffer from issues such as error propagation, massive parameters, and difficulties in deployment and retaining visual characteristics of the input image. Thus, constructing end-to-end models has become an option, which, however, faces two main challenges: 1) the huge modeling burden, as it is required to simultaneously learn alignment across languages and preserve the visual characteristics of the input image; 2) the difficulties of directly predicting excessively lengthy pixel sequences. In this paper, we propose \textit{Translatotron-V(ision)}, an end-to-end IIMT model consisting of four modules. In addition to an image encoder, and an image decoder, our model contains a target text decoder and an image tokenizer. Among them, the target text decoder is used to alleviate the language alignment burden, and the image tokenizer converts long sequences of pixels into shorter sequences of visual tokens, preventing the model from focusing on low-level visual features. Besides, we present a two-stage training framework for our model to assist the model in learning alignment across modalities and languages. Finally, we propose a location-aware evaluation metric called Structure-BLEU to assess the translation quality of the generated images. Experimental results demonstrate that our model achieves competitive performance compared to cascaded models with only 70.9\% of parameters, and significantly outperforms the pixel-level end-to-end IIMT model.

WeLayout: WeChat Layout Analysis System for the ICDAR 2023 Competition on Robust Layout Segmentation in Corporate Documents

In this paper, we introduce WeLayout, a novel system for segmenting the layout of corporate documents, which stands for WeChat Layout Analysis System. Our approach utilizes a sophisticated ensemble of DINO and YOLO models, specifically developed for the ICDAR 2023 Competition on Robust Layout Segmentation. Our method significantly surpasses the baseline, securing a top position on the leaderboard with a mAP of 70.0. To achieve this performance, we concentrated on enhancing various aspects of the task, such as dataset augmentation, model architecture, bounding box refinement, and model ensemble techniques. Additionally, we trained the data separately for each document category to ensure a higher mean submission score. We also developed an algorithm for cell matching to further improve our performance. To identify the optimal weights and IoU thresholds for our model ensemble, we employed a Bayesian optimization algorithm called the Tree-Structured Parzen Estimator. Our approach effectively demonstrates the benefits of combining query-based and anchor-free models for achieving robust layout segmentation in corporate documents.