To improve instance-level detection/segmentation performance, existing self-supervised and semi-supervised methods extract either very task-unrelated or very task-specific training signals from unlabeled data. We argue that these two approaches, at the two extreme ends of the task-specificity spectrum, are suboptimal for the task performance. Utilizing too little task-specific training signals causes underfitting to the ground-truth labels of downstream tasks, while the opposite causes overfitting to the ground-truth labels. To this end, we propose a novel Class-agnostic Semi-supervised Pretraining (CaSP) framework to achieve a more favorable task-specificity balance in extracting training signals from unlabeled data. Compared to semi-supervised learning, CaSP reduces the task specificity in training signals by ignoring class information in the pseudo labels and having a separate pretraining stage that uses only task-unrelated unlabeled data. On the other hand, CaSP preserves the right amount of task specificity by leveraging box/mask-level pseudo labels. As a result, our pretrained model can better avoid underfitting/overfitting to ground-truth labels when finetuned on the downstream task. Using 3.6M unlabeled data, we achieve a remarkable performance gain of 4.7% over ImageNet-pretrained baseline on object detection. Our pretrained model also demonstrates excellent transferability to other detection and segmentation tasks/frameworks.
One of the major challenges in training text-to-image generation models is the need of a large number of high-quality image-text pairs. While image samples are often easily accessible, the associated text descriptions typically require careful human captioning, which is particularly time- and cost-consuming. In this paper, we propose the first work to train text-to-image generation models without any text data. Our method leverages the well-aligned multi-modal semantic space of the powerful pre-trained CLIP model: the requirement of text-conditioning is seamlessly alleviated via generating text features from image features. Extensive experiments are conducted to illustrate the effectiveness of the proposed method. We obtain state-of-the-art results in the standard text-to-image generation tasks. Importantly, the proposed language-free model outperforms most existing models trained with full image-text pairs. Furthermore, our method can be applied in fine-tuning pre-trained models, which saves both training time and cost in training text-to-image generation models. Our pre-trained model obtains competitive results in zero-shot text-to-image generation on the MS-COCO dataset, yet with around only 1% of the model size and training data size relative to the recently proposed large DALL-E model.
Open-vocabulary instance segmentation aims at segmenting novel classes without mask annotations. It is an important step toward reducing laborious human supervision. Most existing works first pretrain a model on captioned images covering many novel classes and then finetune it on limited base classes with mask annotations. However, the high-level textual information learned from caption pretraining alone cannot effectively encode the details required for pixel-wise segmentation. To address this, we propose a cross-modal pseudo-labeling framework, which generates training pseudo masks by aligning word semantics in captions with visual features of object masks in images. Thus, our framework is capable of labeling novel classes in captions via their word semantics to self-train a student model. To account for noises in pseudo masks, we design a robust student model that selectively distills mask knowledge by estimating the mask noise levels, hence mitigating the adverse impact of noisy pseudo masks. By extensive experiments, we show the effectiveness of our framework, where we significantly improve mAP score by 4.5% on MS-COCO and 5.1% on the large-scale Open Images & Conceptual Captions datasets compared to the state-of-the-art.
In instance-level detection tasks (e.g., object detection), reducing input resolution is an easy option to improve runtime efficiency. However, this option traditionally hurts the detection performance much. This paper focuses on boosting the performance of low-resolution models by distilling knowledge from a high- or multi-resolution model. We first identify the challenge of applying knowledge distillation (KD) to teacher and student networks that act on different input resolutions. To tackle it, we explore the idea of spatially aligning feature maps between models of varying input resolutions by shifting feature pyramid positions and introduce aligned multi-scale training to train a multi-scale teacher that can distill its knowledge to a low-resolution student. Further, we propose crossing feature-level fusion to dynamically fuse teacher's multi-resolution features to guide the student better. On several instance-level detection tasks and datasets, the low-resolution models trained via our approach perform competitively with high-resolution models trained via conventional multi-scale training, while outperforming the latter's low-resolution models by 2.1% to 3.6% in terms of mAP. Our code is made publicly available at https://github.com/dvlab-research/MSAD.
We introduce a new image segmentation task, termed Entity Segmentation (ES) with the aim to segment all visual entities in an image without considering semantic category labels. It has many practical applications in image manipulation/editing where the segmentation mask quality is typically crucial but category labels are less important. In this setting, all semantically-meaningful segments are equally treated as categoryless entities and there is no thing-stuff distinction. Based on our unified entity representation, we propose a center-based entity segmentation framework with two novel modules to improve mask quality. Experimentally, both our new task and framework demonstrate superior advantages as against existing work. In particular, ES enables the following: (1) merging multiple datasets to form a large training set without the need to resolve label conflicts; (2) any model trained on one dataset can generalize exceptionally well to other datasets with unseen domains. Our code is made publicly available at https://github.com/dvlab-research/Entity.
We propose SelfDoc, a task-agnostic pre-training framework for document image understanding. Because documents are multimodal and are intended for sequential reading, our framework exploits the positional, textual, and visual information of every semantically meaningful component in a document, and it models the contextualization between each block of content. Unlike existing document pre-training models, our model is coarse-grained instead of treating individual words as input, therefore avoiding an overly fine-grained with excessive contextualization. Beyond that, we introduce cross-modal learning in the model pre-training phase to fully leverage multimodal information from unlabeled documents. For downstream usage, we propose a novel modality-adaptive attention mechanism for multimodal feature fusion by adaptively emphasizing language and vision signals. Our framework benefits from self-supervised pre-training on documents without requiring annotations by a feature masking training strategy. It achieves superior performance on multiple downstream tasks with significantly fewer document images used in the pre-training stage compared to previous works.
Recent studies indicate that NLU models are prone to rely on shortcut features for prediction, without achieving true language understanding. As a result, these models fail to generalize to real-world out-of-distribution data. In this work, we show that the words in the NLU training set can be modeled as a long-tailed distribution. There are two findings: 1) NLU models have strong preference for features located at the head of the long-tailed distribution, and 2) Shortcut features are picked up during very early few iterations of the model training. These two observations are further employed to formulate a measurement which can quantify the shortcut degree of each training sample. Based on this shortcut measurement, we propose a shortcut mitigation framework LGTR, to suppress the model from making overconfident predictions for samples with large shortcut degree. Experimental results on three NLU benchmarks demonstrate that our long-tailed distribution explanation accurately reflects the shortcut learning behavior of NLU models. Experimental analysis further indicates that LGTR can improve the generalization accuracy on OOD data, while preserving the accuracy on in-distribution data.
Most recent image captioning works are conducted in English as the majority of image-caption datasets are in English. However, there are a large amount of non-native English speakers worldwide. Generating image captions in different languages is worth exploring. In this paper, we present a novel unsupervised method to generate image captions without using any caption corpus. Our method relies on 1) a cross-lingual auto-encoding, which learns the scene graph mapping function along with the scene graph encoders and sentence decoders on machine translation parallel corpora, and 2) an unsupervised feature mapping, which seeks to map the encoded scene graph features from image modality to sentence modality. By leveraging cross-lingual auto-encoding, cross-modal feature mapping, and adversarial learning, our method can learn an image captioner to generate captions in different languages. We verify the effectiveness of our proposed method on the Chinese image caption generation. The comparisons against several baseline methods demonstrate the effectiveness of our approach.
Change Captioning is a task that aims to describe the difference between images with natural language. Most existing methods treat this problem as a difference judgment without the existence of distractors, such as viewpoint changes. However, in practice, viewpoint changes happen often and can overwhelm the semantic difference to be described. In this paper, we propose a novel visual encoder to explicitly distinguish viewpoint changes from semantic changes in the change captioning task. Moreover, we further simulate the attention preference of humans and propose a novel reinforcement learning process to fine-tune the attention directly with language evaluation rewards. Extensive experimental results show that our method outperforms the state-of-the-art approaches by a large margin in both Spot-the-Diff and CLEVR-Change datasets.