Token interaction operation is one of the core modules in MLP-based models to exchange and aggregate information between different spatial locations. However, the power of token interaction on the spatial dimension is highly dependent on the spatial resolution of the feature maps, which limits the model's expressive ability, especially in deep layers where the feature are down-sampled to a small spatial size. To address this issue, we present a novel method called \textbf{Strip-MLP} to enrich the token interaction power in three ways. Firstly, we introduce a new MLP paradigm called Strip MLP layer that allows the token to interact with other tokens in a cross-strip manner, enabling the tokens in a row (or column) to contribute to the information aggregations in adjacent but different strips of rows (or columns). Secondly, a \textbf{C}ascade \textbf{G}roup \textbf{S}trip \textbf{M}ixing \textbf{M}odule (CGSMM) is proposed to overcome the performance degradation caused by small spatial feature size. The module allows tokens to interact more effectively in the manners of within-patch and cross-patch, which is independent to the feature spatial size. Finally, based on the Strip MLP layer, we propose a novel \textbf{L}ocal \textbf{S}trip \textbf{M}ixing \textbf{M}odule (LSMM) to boost the token interaction power in the local region. Extensive experiments demonstrate that Strip-MLP significantly improves the performance of MLP-based models on small datasets and obtains comparable or even better results on ImageNet. In particular, Strip-MLP models achieve higher average Top-1 accuracy than existing MLP-based models by +2.44\% on Caltech-101 and +2.16\% on CIFAR-100. The source codes will be available at~\href{https://github.com/Med-Process/Strip_MLP{https://github.com/Med-Process/Strip\_MLP}.
Unpaired Medical Image Enhancement (UMIE) aims to transform a low-quality (LQ) medical image into a high-quality (HQ) one without relying on paired images for training. While most existing approaches are based on Pix2Pix/CycleGAN and are effective to some extent, they fail to explicitly use HQ information to guide the enhancement process, which can lead to undesired artifacts and structural distortions. In this paper, we propose a novel UMIE approach that avoids the above limitation of existing methods by directly encoding HQ cues into the LQ enhancement process in a variational fashion and thus model the UMIE task under the joint distribution between the LQ and HQ domains. Specifically, we extract features from an HQ image and explicitly insert the features, which are expected to encode HQ cues, into the enhancement network to guide the LQ enhancement with the variational normalization module. We train the enhancement network adversarially with a discriminator to ensure the generated HQ image falls into the HQ domain. We further propose a content-aware loss to guide the enhancement process with wavelet-based pixel-level and multi-encoder-based feature-level constraints. Additionally, as a key motivation for performing image enhancement is to make the enhanced images serve better for downstream tasks, we propose a bi-level learning scheme to optimize the UMIE task and downstream tasks cooperatively, helping generate HQ images both visually appealing and favorable for downstream tasks. Experiments on three medical datasets, including two newly collected datasets, verify that the proposed method outperforms existing techniques in terms of both enhancement quality and downstream task performance. We will make the code and the newly collected datasets publicly available for community study.
Mixup style data augmentation algorithms have been widely adopted in various tasks as implicit network regularization on representation learning to improve model generalization, which can be achieved by a linear interpolation of labeled samples in input or feature space as well as target space. Inspired by good robustness of alternative dropout strategies against over-fitting on limited patterns of training samples, this paper introduces a novel concept of ShuffleMix -- Shuffle of Mixed hidden features, which can be interpreted as a kind of dropout operation in feature space. Specifically, our ShuffleMix method favors a simple linear shuffle of randomly selected feature channels for feature mixup in-between training samples to leverage semantic interpolated supervision signals, which can be extended to a generalized shuffle operation via additionally combining linear interpolations of intra-channel features. Compared to its direct competitor of feature augmentation -- the Manifold Mixup, the proposed ShuffleMix can gain superior generalization, owing to imposing more flexible and smooth constraints on generating samples and achieving regularization effects of channel-wise feature dropout. Experimental results on several public benchmarking datasets of single-label and multi-label visual classification tasks can confirm the effectiveness of our method on consistently improving representations over the state-of-the-art mixup augmentation.
Deep representation learning is a subfield of machine learning that focuses on learning meaningful and useful representations of data through deep neural networks. However, existing methods for semantic classification typically employ pre-defined target codes such as the one-hot and the Hadamard codes, which can either fail or be less flexible to model inter-class correlation. In light of this, this paper introduces a novel learnable target coding as an auxiliary regularization of deep representation learning, which can not only incorporate latent dependency across classes but also impose geometric properties of target codes into representation space. Specifically, a margin-based triplet loss and a correlation consistency loss on the proposed target codes are designed to encourage more discriminative representations owing to enlarging between-class margins in representation space and favoring equal semantic correlation of learnable target codes respectively. Experimental results on several popular visual classification and retrieval benchmarks can demonstrate the effectiveness of our method on improving representation learning, especially for imbalanced data.
Domain gap between synthetic and real data in visual regression (\eg 6D pose estimation) is bridged in this paper via global feature alignment and local refinement on the coarse classification of discretized anchor classes in target space, which imposes a piece-wise target manifold regularization into domain-invariant representation learning. Specifically, our method incorporates an explicit self-supervised manifold regularization, revealing consistent cumulative target dependency across domains, to a self-training scheme (\eg the popular Self-Paced Self-Training) to encourage more discriminative transferable representations of regression tasks. Moreover, learning unified implicit neural functions to estimate relative direction and distance of targets to their nearest class bins aims to refine target classification predictions, which can gain robust performance against inconsistent feature scaling sensitive to UDA regressors. Experiment results on three public benchmarks of the challenging 6D pose estimation task can verify the effectiveness of our method, consistently achieving superior performance to the state-of-the-art for UDA on 6D pose estimation.
Visual Grounding (VG) refers to locating a region described by expressions in a specific image, which is a critical topic in vision-language fields. To alleviate the dependence on labeled data, existing unsupervised methods try to locate regions using task-unrelated pseudo-labels. However, a large proportion of pseudo-labels are noisy and diversity scarcity in language taxonomy. Inspired by the advances in V-L pretraining, we consider utilizing the VLP models to realize unsupervised transfer learning in downstream grounding task. Thus, we propose CLIP-VG, a novel method that can conduct self-paced curriculum adapting of CLIP via exploiting pseudo-language labels to solve VG problem. By elaborating an efficient model structure, we first propose a single-source and multi-source curriculum adapting method for unsupervised VG to progressively sample more reliable cross-modal pseudo-labels to obtain the optimal model, thus achieving implicit knowledge exploiting and denoising. Our method outperforms the existing state-of-the-art unsupervised VG method Pseudo-Q in both single-source and multi-source scenarios with a large margin, i.e., 6.78%~10.67% and 11.39%~24.87% on RefCOCO/+/g datasets, even outperforms existing weakly supervised methods. The code and models will be released at \url{https://github.com/linhuixiao/CLIP-VG}.
Vision foundation models exhibit impressive power, benefiting from the extremely large model capacity and broad training data. However, in practice, downstream scenarios may only support a small model due to the limited computational resources or efficiency considerations. Moreover, the data used for pretraining foundation models are usually invisible and very different from the target data of downstream tasks. This brings a critical challenge for the real-world application of foundation models: one has to transfer the knowledge of a foundation model to the downstream task that has a quite different architecture with only downstream target data. Existing transfer learning or knowledge distillation methods depend on either the same model structure or finetuning of the foundation model. Thus, naively introducing these methods can be either infeasible or very inefficient. To address this, we propose a Task-Driven Model Reprogramming (TDMR) framework. Specifically, we reprogram the foundation model to project the knowledge into a proxy space, which alleviates the adverse effect of task mismatch and domain inconsistency. Then, we reprogram the target model via progressive distillation from the proxy space to efficiently learn the knowledge from the reprogrammed foundation model. TDMR is compatible with different pre-trained model types (CNN, transformer or their mix) and limited target data, and promotes the wide applications of vision foundation models to downstream tasks in a cost-effective manner. Extensive experiments on different downstream classification tasks and target model structures demonstrate the effectiveness of our methods with both CNNs and transformer foundation models.
Vision-and-language navigation (VLN) is the task to enable an embodied agent to navigate to a remote location following the natural language instruction in real scenes. Most of the previous approaches utilize the entire features or object-centric features to represent navigable candidates. However, these representations are not efficient enough for an agent to perform actions to arrive the target location. As knowledge provides crucial information which is complementary to visible content, in this paper, we propose a Knowledge Enhanced Reasoning Model (KERM) to leverage knowledge to improve agent navigation ability. Specifically, we first retrieve facts (i.e., knowledge described by language descriptions) for the navigation views based on local regions from the constructed knowledge base. The retrieved facts range from properties of a single object (e.g., color, shape) to relationships between objects (e.g., action, spatial position), providing crucial information for VLN. We further present the KERM which contains the purification, fact-aware interaction, and instruction-guided aggregation modules to integrate visual, history, instruction, and fact features. The proposed KERM can automatically select and gather crucial and relevant cues, obtaining more accurate action prediction. Experimental results on the REVERIE, R2R, and SOON datasets demonstrate the effectiveness of the proposed method.
This paper aims to solve the video object segmentation (VOS) task in a scribble-supervised manner, in which VOS models are not only trained by the sparse scribble annotations but also initialized with the sparse target scribbles for inference. Thus, the annotation burdens for both training and initialization can be substantially lightened. The difficulties of scribble-supervised VOS lie in two aspects. On the one hand, it requires the powerful ability to learn from the sparse scribble annotations during training. On the other hand, it demands strong reasoning capability during inference given only a sparse initial target scribble. In this work, we propose a Reliability-Hierarchical Memory Network (RHMNet) to predict the target mask in a step-wise expanding strategy w.r.t. the memory reliability level. To be specific, RHMNet first only uses the memory in the high-reliability level to locate the region with high reliability belonging to the target, which is highly similar to the initial target scribble. Then it expands the located high-reliability region to the entire target conditioned on the region itself and the memories in all reliability levels. Besides, we propose a scribble-supervised learning mechanism to facilitate the learning of our model to predict dense results. It mines the pixel-level relation within the single frame and the frame-level relation within the sequence to take full advantage of the scribble annotations in sequence training samples. The favorable performance on two popular benchmarks demonstrates that our method is promising.
Natural Language Generation (NLG) accepts input data in the form of images, videos, or text and generates corresponding natural language text as output. Existing NLG methods mainly adopt a supervised approach and rely heavily on coupled data-to-text pairs. However, for many targeted scenarios and for non-English languages, sufficient quantities of labeled data are often not available. To relax the dependency on labeled data of downstream tasks, we propose an intuitive and effective zero-shot learning framework, ZeroNLG, which can deal with multiple NLG tasks, including image-to-text (image captioning), video-to-text (video captioning), and text-to-text (neural machine translation), across English, Chinese, German, and French within a unified framework. ZeroNLG does not require any labeled downstream pairs for training. During training, ZeroNLG (i) projects different domains (across modalities and languages) to corresponding coordinates in a shared common latent space; (ii) bridges different domains by aligning their corresponding coordinates in this space; and (iii) builds an unsupervised multilingual auto-encoder to learn to generate text by reconstructing the input text given its coordinate in shared latent space. Consequently, during inference, based on the data-to-text pipeline, ZeroNLG can generate target sentences across different languages given the coordinate of input data in the common space. Within this unified framework, given visual (imaging or video) data as input, ZeroNLG can perform zero-shot visual captioning; given textual sentences as input, ZeroNLG can perform zero-shot machine translation. We present the results of extensive experiments on twelve NLG tasks, showing that, without using any labeled downstream pairs for training, ZeroNLG generates high-quality and believable outputs and significantly outperforms existing zero-shot methods.