No abstract

Unsupervised large-scale vision-language pre-training has shown promising advances on various downstream tasks. Existing methods often model the crossmodal interaction either via the similarity of the global feature of each modality which misses sufficient information, or finer-grained interactions using cross/selfattention upon visual and textual tokens. However, cross/self-attention suffers from inferior efficiency in both training and inference. In this paper, we introduce a large-scale Fine-grained Interactive Language-Image Pre-training (FILIP) to achieve finer-level alignment through a cross-modal late interaction mechanism, which uses a token-wise maximum similarity between visual and textual tokens to guide the contrastive objective. FILIP successfully leverages the finergrained expressiveness between image patches and textual words by modifying only contrastive loss, while simultaneously gaining the ability to pre-compute image and text representations offline at inference, keeping both large-scale training and inference efficient. Furthermore, we construct a new large-scale image-text pair dataset called FILIP300M for pre-training. Experiments show that FILIP achieves state-of-the-art performance on multiple downstream vision-language tasks including zero-shot image classification and image-text retrieval. The visualization on word-patch alignment further shows that FILIP can learn meaningful fine-grained features with promising localization ability.

The state-of-the-art object detection method is complicated with various modules such as backbone, RPN, feature fusion neck and RCNN head, where each module may have different designs and structures. How to leverage the computational cost and accuracy trade-off for the structural combination as well as the modular selection of multiple modules? Neural architecture search (NAS) has shown great potential in finding an optimal solution. Existing NAS works for object detection only focus on searching better design of a single module such as backbone or feature fusion neck, while neglecting the balance of the whole system. In this paper, we present a two-stage coarse-to-fine searching strategy named Structural-to-Modular NAS (SM-NAS) for searching a GPU-friendly design of both an efficient combination of modules and better modular-level architecture for object detection. Specifically, Structural-level searching stage first aims to find an efficient combination of different modules; Modular-level searching stage then evolves each specific module and pushes the Pareto front forward to a faster task-specific network. We consider a multi-objective search where the search space covers many popular designs of detection methods. We directly search a detection backbone without pre-trained models or any proxy task by exploring a fast training from scratch strategy. The resulting architectures dominate state-of-the-art object detection systems in both inference time and accuracy and demonstrate the effectiveness on multiple detection datasets, e.g. halving the inference time with additional 1% mAP improvement compared to FPN and reaching 46% mAP with the similar inference time of MaskRCNN.

This paper presents a large-scale Chinese cross-modal dataset for benchmarking different multi-modal pre-training methods to facilitate the Vision-Language Pretraining (VLP) research and community development. Recent dual-stream VLP models like CLIP, ALIGN and FILIP have shown remarkable performance on various downstream tasks as well as their remarkable zero-shot ability in the open domain tasks. However, their success heavily relies on the scale of pre-trained datasets. Though there have been both small-scale vision-language English datasets like Flickr30k, CC12M as well as large-scale LAION-400M, the current community lacks large-scale Vision-Language benchmarks in Chinese, hindering the development of broader multilingual applications. On the other hand, there is very rare publicly available large-scale Chinese cross-modal pre-training dataset that has been released, making it hard to use pre-trained models as services for downstream tasks. In this work, we release a Large-Scale Chinese Cross-modal dataset named Wukong, containing 100 million Chinese image-text pairs from the web. Furthermore, we release a group of big models pre-trained with advanced image encoders (ResNet/ViT/SwinT) and different pre-training methods (CLIP/FILIP/LiT). We provide extensive experiments, a deep benchmarking of different downstream tasks, and some exciting findings. Experiments show that Wukong can serve as a promising Chinese pre-training dataset and benchmark for different cross-modal learning methods, which gives superior performance on various downstream tasks such as zero-shot image classification and image-text retrieval benchmarks. More information can refer to https://wukong-dataset.github.io/wukong-dataset/.

The state-of-the-art object detection method is complicated with various modules such as backbone, feature fusion neck, RPN, and RCNN head, where each module may have different designs and structures. How to leverage the computational cost and accuracy trade-off for the structural combination as well as the modular selection of multiple modules? Neural architecture search (NAS) has shown great potential in finding an optimal solution. Existing NAS works for object detection only focus on searching better design of a single module such as backbone or feature fusion neck, while neglecting the balance of the whole system. In this paper, we present a two-stage coarse-to-fine searching strategy named Structural-to-Modular NAS (SM-NAS) for searching a GPU-friendly design of both an efficient combination of modules and better modular-level architecture for object detection. Specifically, Structural-level searching stage first aims to find an efficient combination of different modules; Modular-level searching stage then evolves each specific module and pushes the Pareto front forward to a faster task-specific network. We consider a multi-objective search where the search space covers many popular designs of detection methods. We directly search a detection backbone without pre-trained models or any proxy task by exploring a fast training from scratch strategy. The resulting architectures dominate state-of-the-art object detection systems in both inference time and accuracy and demonstrate the effectiveness on multiple detection datasets, e.g. halving the inference time with additional 1% mAP improvement compared to FPN and reaching 46% mAP with the similar inference time of MaskRCNN.