One Transformer Can Understand Both 2D & 3D Molecular Data

Unlike vision and language data which usually has a unique format, molecules can naturally be characterized using different chemical formulations. One can view a molecule as a 2D graph or define it as a collection of atoms located in a 3D space. For molecular representation learning, most previous works designed neural networks only for a particular data format, making the learned models likely to fail for other data formats. We believe a general-purpose neural network model for chemistry should be able to handle molecular tasks across data modalities. To achieve this goal, in this work, we develop a novel Transformer-based Molecular model called Transformer-M, which can take molecular data of 2D or 3D formats as input and generate meaningful semantic representations. Using the standard Transformer as the backbone architecture, Transformer-M develops two separated channels to encode 2D and 3D structural information and incorporate them with the atom features in the network modules. When the input data is in a particular format, the corresponding channel will be activated, and the other will be disabled. By training on 2D and 3D molecular data with properly designed supervised signals, Transformer-M automatically learns to leverage knowledge from different data modalities and correctly capture the representations. We conducted extensive experiments for Transformer-M. All empirical results show that Transformer-M can simultaneously achieve strong performance on 2D and 3D tasks, suggesting its broad applicability. The code and models will be made publicly available at https://github.com/lsj2408/Transformer-M.

TransVG++: End-to-End Visual Grounding with Language Conditioned Vision Transformer

In this work, we explore neat yet effective Transformer-based frameworks for visual grounding. The previous methods generally address the core problem of visual grounding, i.e., multi-modal fusion and reasoning, with manually-designed mechanisms. Such heuristic designs are not only complicated but also make models easily overfit specific data distributions. To avoid this, we first propose TransVG, which establishes multi-modal correspondences by Transformers and localizes referred regions by directly regressing box coordinates. We empirically show that complicated fusion modules can be replaced by a simple stack of Transformer encoder layers with higher performance. However, the core fusion Transformer in TransVG is stand-alone against uni-modal encoders, and thus should be trained from scratch on limited visual grounding data, which makes it hard to be optimized and leads to sub-optimal performance. To this end, we further introduce TransVG++ to make two-fold improvements. For one thing, we upgrade our framework to a purely Transformer-based one by leveraging Vision Transformer (ViT) for vision feature encoding. For another, we devise Language Conditioned Vision Transformer that removes external fusion modules and reuses the uni-modal ViT for vision-language fusion at the intermediate layers. We conduct extensive experiments on five prevalent datasets, and report a series of state-of-the-art records.

TransVG: End-to-End Visual Grounding with Transformers

In this paper, we present a neat yet effective transformer-based framework for visual grounding, namely TransVG, to address the task of grounding a language query to the corresponding region onto an image. The state-of-the-art methods, including two-stage or one-stage ones, rely on a complex module with manually-designed mechanisms to perform the query reasoning and multi-modal fusion. However, the involvement of certain mechanisms in fusion module design, such as query decomposition and image scene graph, makes the models easily overfit to datasets with specific scenarios, and limits the plenitudinous interaction between the visual-linguistic context. To avoid this caveat, we propose to establish the multi-modal correspondence by leveraging transformers, and empirically show that the complex fusion modules (e.g., modular attention network, dynamic graph, and multi-modal tree) can be replaced by a simple stack of transformer encoder layers with higher performance. Moreover, we re-formulate the visual grounding as a direct coordinates regression problem and avoid making predictions out of a set of candidates (i.e., region proposals or anchor boxes). Extensive experiments are conducted on five widely used datasets, and a series of state-of-the-art records are set by our TransVG. We build the benchmark of transformer-based visual grounding framework and will make our code available to the public.

Improving One-stage Visual Grounding by Recursive Sub-query Construction

We improve one-stage visual grounding by addressing current limitations on grounding long and complex queries. Existing one-stage methods encode the entire language query as a single sentence embedding vector, e.g., taking the embedding from BERT or the hidden state from LSTM. This single vector representation is prone to overlooking the detailed descriptions in the query. To address this query modeling deficiency, we propose a recursive sub-query construction framework, which reasons between image and query for multiple rounds and reduces the referring ambiguity step by step. We show our new one-stage method obtains 5.0%, 4.5%, 7.5%, 12.8% absolute improvements over the state-of-the-art one-stage baseline on ReferItGame, RefCOCO, RefCOCO+, and RefCOCOg, respectively. In particular, superior performances on longer and more complex queries validates the effectiveness of our query modeling.

Global Image Sentiment Transfer

Transferring the sentiment of an image is an unexplored research topic in the area of computer vision. This work proposes a novel framework consisting of a reference image retrieval step and a global sentiment transfer step to transfer sentiments of images according to a given sentiment tag. The proposed image retrieval algorithm is based on the SSIM index. The retrieved reference images by the proposed algorithm are more content-related against the algorithm based on the perceptual loss. Therefore can lead to a better image sentiment transfer result. In addition, we propose a global sentiment transfer step, which employs an optimization algorithm to iteratively transfer sentiment of images based on feature maps produced by the Densenet121 architecture. The proposed sentiment transfer algorithm can transfer the sentiment of images while ensuring the content structure of the input image intact. The qualitative and quantitative experiments demonstrate that the proposed sentiment transfer framework outperforms existing artistic and photorealistic style transfer algorithms in making reliable sentiment transfer results with rich and exact details.

Image Sentiment Transfer

In this work, we introduce an important but still unexplored research task -- image sentiment transfer. Compared with other related tasks that have been well-studied, such as image-to-image translation and image style transfer, transferring the sentiment of an image is more challenging. Given an input image, the rule to transfer the sentiment of each contained object can be completely different, making existing approaches that perform global image transfer by a single reference image inadequate to achieve satisfactory performance. In this paper, we propose an effective and flexible framework that performs image sentiment transfer at the object level. It first detects the objects and extracts their pixel-level masks, and then performs object-level sentiment transfer guided by multiple reference images for the corresponding objects. For the core object-level sentiment transfer, we propose a novel Sentiment-aware GAN (SentiGAN). Both global image-level and local object-level supervisions are imposed to train SentiGAN. More importantly, an effective content disentanglement loss cooperating with a content alignment step is applied to better disentangle the residual sentiment-related information of the input image. Extensive quantitative and qualitative experiments are performed on the object-oriented VSO dataset we create, demonstrating the effectiveness of the proposed framework.

Example-Guided Image Synthesis across Arbitrary Scenes using Masked Spatial-Channel Attention and Self-Supervision

Example-guided image synthesis has recently been attempted to synthesize an image from a semantic label map and an exemplary image. In the task, the additional exemplar image provides the style guidance that controls the appearance of the synthesized output. Despite the controllability advantage, the existing models are designed on datasets with specific and roughly aligned objects. In this paper, we tackle a more challenging and general task, where the exemplar is an arbitrary scene image that is semantically different from the given label map. To this end, we first propose a Masked Spatial-Channel Attention (MSCA) module which models the correspondence between two arbitrary scenes via efficient decoupled attention. Next, we propose an end-to-end network for joint global and local feature alignment and synthesis. Finally, we propose a novel self-supervision task to enable training. Experiments on the large-scale and more diverse COCO-stuff dataset show significant improvements over the existing methods. Moreover, our approach provides interpretability and can be readily extended to other content manipulation tasks including style and spatial interpolation or extrapolation.

Anatomy-aware 3D Human Pose Estimation in Videos

In this work, we propose a new solution for 3D human pose estimation in videos. Instead of directly regressing the 3D joint locations, we draw inspiration from the human skeleton anatomy and decompose the task into bone direction prediction and bone length prediction, from which the 3D joint locations can be completely derived. Our motivation is the fact that the bone lengths of a human skeleton remain consistent across time. This promotes us to develop effective techniques to utilize global information across {\it all} the frames in a video for high-accuracy bone length prediction. Moreover, for the bone direction prediction network, we propose a fully-convolutional propagating architecture with long skip connections. Essentially, it predicts the directions of different bones hierarchically without using any time-consuming memory units (e.g. LSTM). A novel joint shift loss is further introduced to bridge the training of the bone length and bone direction prediction networks. Finally, we employ an implicit attention mechanism to feed the 2D keypoint visibility scores into the model as extra guidance, which significantly mitigates the depth ambiguity in many challenging poses. Our full model outperforms the previous best results on Human3.6M and MPI-INF-3DHP datasets, where comprehensive evaluation validates the effectiveness of our model.

Adaptive Offline Quintuplet Loss for Image-Text Matching

Existing image-text matching approaches typically leverage triplet loss with online hard negatives to train the model. For each image or text anchor in a training mini-batch, the model is trained to distinguish between a positive and the most confusing negative of the anchor mined from the mini-batch (i.e. online hard negative). This strategy improves the model's capacity to discover fine-grained correspondences and non-correspondences between image and text inputs. However, the above training approach has the following drawbacks: (1) the negative selection strategy still provides limited chances for the model to learn from very hard-to-distinguish cases. (2) The trained model has weak generalization capability from the training set to the testing set. (3) The penalty lacks hierarchy and adaptiveness for hard negatives with different ``hardness'' degrees. In this paper, we propose solutions by sampling negatives offline from the whole training set. It provides ``harder'' offline negatives than online hard negatives for the model to distinguish. Based on the offline hard negatives, a quintuplet loss is proposed to improve the model's generalization capability to distinguish positives and negatives. In addition, a novel loss function that combines the knowledge of positives, offline hard negatives and online hard negatives is created. It leverages offline hard negatives as intermediary to adaptively penalize them based on their distance relations to the anchor. We evaluate the proposed training approach on three state-of-the-art image-text models on the MS-COCO and Flickr30K datasets. Significant performance improvements are observed for all the models, demonstrating the effectiveness and generality of the proposed approach.