Modern lens designs are capable of resolving >10 gigapixels, while advances in camera frame-rate and hyperspectral imaging have made Terapixel/s data acquisition a real possibility. The main bottlenecks preventing such high data-rate systems are power consumption and data storage. In this work, we show that analog photonic encoders could address this challenge, enabling high-speed image compression using orders-of-magnitude lower power than digital electronics. Our approach relies on a silicon-photonics front-end to compress raw image data, foregoing energy-intensive image conditioning and reducing data storage requirements. The compression scheme uses a passive disordered photonic structure to perform kernel-type random projections of the raw image data with minimal power consumption and low latency. A back-end neural network can then reconstruct the original images with structural similarity exceeding 90%. This scheme has the potential to process Terapixel/s data streams using less than 100 fJ/pixel, providing a path to ultra-high-resolution data and image acquisition systems.
Learning based feature matching methods have been commonly studied in recent years. The core issue for learning feature matching is to how to learn (1) discriminative representations for feature points (or regions) within each intra-image and (2) consensus representations for feature points across inter-images. Recently, self- and cross-attention models have been exploited to address this issue. However, in many scenes, features are coming with large-scale, redundant and outliers contaminated. Previous self-/cross-attention models generally conduct message passing on all primal features which thus lead to redundant learning and high computational cost. To mitigate limitations, inspired by recent seed matching methods, in this paper, we propose a novel efficient Anchor Matching Transformer (AMatFormer) for the feature matching problem. AMatFormer has two main aspects: First, it mainly conducts self-/cross-attention on some anchor features and leverages these anchor features as message bottleneck to learn the representations for all primal features. Thus, it can be implemented efficiently and compactly. Second, AMatFormer adopts a shared FFN module to further embed the features of two images into the common domain and thus learn the consensus feature representations for the matching problem. Experiments on several benchmarks demonstrate the effectiveness and efficiency of the proposed AMatFormer matching approach.
We present the training recipe and results of scaling up PaLI-X, a multilingual vision and language model, both in terms of size of the components and the breadth of its training task mixture. Our model achieves new levels of performance on a wide-range of varied and complex tasks, including multiple image-based captioning and question-answering tasks, image-based document understanding and few-shot (in-context) learning, as well as object detection, video question answering, and video captioning. PaLI-X advances the state-of-the-art on most vision-and-language benchmarks considered (25+ of them). Finally, we observe emerging capabilities, such as complex counting and multilingual object detection, tasks that are not explicitly in the training mix.
We introduce Three Towers (3T), a flexible method to improve the contrastive learning of vision-language models by incorporating pretrained image classifiers. While contrastive models are usually trained from scratch, LiT (Zhai et al., 2022) has recently shown performance gains from using pretrained classifier embeddings. However, LiT directly replaces the image tower with the frozen embeddings, excluding any potential benefits of contrastively training the image tower. With 3T, we propose a more flexible strategy that allows the image tower to benefit from both pretrained embeddings and contrastive training. To achieve this, we introduce a third tower that contains the frozen pretrained embeddings, and we encourage alignment between this third tower and the main image-text towers. Empirically, 3T consistently improves over LiT and the CLIP-style from-scratch baseline for retrieval tasks. For classification, 3T reliably improves over the from-scratch baseline, and while it underperforms relative to LiT for JFT-pretrained models, it outperforms LiT for ImageNet-21k and Places365 pretraining.
Pretrained language models have achieved remarkable success in various natural language processing tasks. However, pretraining has recently shifted toward larger models and larger data, and this has resulted in significant computational and energy costs. In this paper, we propose Influence Subset Selection (ISS) for language model, which explicitly utilizes end-task knowledge to select a tiny subset of the pretraining corpus. Specifically, the ISS selects the samples that will provide the most positive influence on the performance of the end-task. Furthermore, we design a gradient matching based influence estimation method, which can drastically reduce the computation time of influence. With only 0.45% of the data and a three-orders-of-magnitude lower computational cost, ISS outperformed pretrained models (e.g., RoBERTa) on eight datasets covering four domains.
Existing models for named entity recognition (NER) are mainly based on large-scale labeled datasets, which always obtain using crowdsourcing. However, it is hard to obtain a unified and correct label via majority voting from multiple annotators for NER due to the large labeling space and complexity of this task. To address this problem, we aim to utilize the original multi-annotator labels directly. Particularly, we propose a Confidence-based Partial Label Learning (CPLL) method to integrate the prior confidence (given by annotators) and posterior confidences (learned by models) for crowd-annotated NER. This model learns a token- and content-dependent confidence via an Expectation-Maximization (EM) algorithm by minimizing empirical risk. The true posterior estimator and confidence estimator perform iteratively to update the true posterior and confidence respectively. We conduct extensive experimental results on both real-world and synthetic datasets, which show that our model can improve performance effectively compared with strong baselines.
Large-scale text-to-image models have demonstrated amazing ability to synthesize diverse and high-fidelity images. However, these models are often violated by several limitations. Firstly, they require the user to provide precise and contextually relevant descriptions for the desired image modifications. Secondly, current models can impose significant changes to the original image content during the editing process. In this paper, we explore ReGeneration learning in an image-to-image Diffusion model (ReDiffuser), that preserves the content of the original image without human prompting and the requisite editing direction is automatically discovered within the text embedding space. To ensure consistent preservation of the shape during image editing, we propose cross-attention guidance based on regeneration learning. This novel approach allows for enhanced expression of the target domain features while preserving the original shape of the image. In addition, we introduce a cooperative update strategy, which allows for efficient preservation of the original shape of an image, thereby improving the quality and consistency of shape preservation throughout the editing process. Our proposed method leverages an existing pre-trained text-image diffusion model without any additional training. Extensive experiments show that the proposed method outperforms existing work in both real and synthetic image editing.
Heterogeneous graph neural networks (HGNNs) as an emerging technique have shown superior capacity of dealing with heterogeneous information network (HIN). However, most HGNNs follow a semi-supervised learning manner, which notably limits their wide use in reality since labels are usually scarce in real applications. Recently, contrastive learning, a self-supervised method, becomes one of the most exciting learning paradigms and shows great potential when there are no labels. In this paper, we study the problem of self-supervised HGNNs and propose a novel co-contrastive learning mechanism for HGNNs, named HeCo. Different from traditional contrastive learning which only focuses on contrasting positive and negative samples, HeCo employs cross-view contrastive mechanism. Specifically, two views of a HIN (network schema and meta-path views) are proposed to learn node embeddings, so as to capture both of local and high-order structures simultaneously. Then the cross-view contrastive learning, as well as a view mask mechanism, is proposed, which is able to extract the positive and negative embeddings from two views. This enables the two views to collaboratively supervise each other and finally learn high-level node embeddings. Moreover, to further boost the performance of HeCo, two additional methods are designed to generate harder negative samples with high quality. Besides the invariant factors, view-specific factors complementally provide the diverse structure information between different nodes, which also should be contained into the final embeddings. Therefore, we need to further explore each view independently and propose a modified model, called HeCo++. Specifically, HeCo++ conducts hierarchical contrastive learning, including cross-view and intra-view contrasts, which aims to enhance the mining of respective structures.
Existing pedestrian attribute recognition (PAR) algorithms are mainly developed based on a static image. However, the performance is not reliable for images with challenging factors, such as heavy occlusion, motion blur, etc. In this work, we propose to understand human attributes using video frames that can make full use of temporal information. Specifically, we formulate the video-based PAR as a vision-language fusion problem and adopt pre-trained big models CLIP to extract the feature embeddings of given video frames. To better utilize the semantic information, we take the attribute list as another input and transform the attribute words/phrase into the corresponding sentence via split, expand, and prompt. Then, the text encoder of CLIP is utilized for language embedding. The averaged visual tokens and text tokens are concatenated and fed into a fusion Transformer for multi-modal interactive learning. The enhanced tokens will be fed into a classification head for pedestrian attribute prediction. Extensive experiments on a large-scale video-based PAR dataset fully validated the effectiveness of our proposed framework.