Zihao Wang

Zihao Wang, Shaofei Cai, Anji Liu et al. · pku

Achieving human-like planning and control with multimodal observations in an open world is a key milestone for more functional generalist agents. Existing approaches can handle certain long-horizon tasks in an open world. However, they still struggle when the number of open-world tasks could potentially be infinite and lack the capability to progressively enhance task completion as game time progresses. We introduce JARVIS-1, an open-world agent that can perceive multimodal input (visual observations and human instructions), generate sophisticated plans, and perform embodied control, all within the popular yet challenging open-world Minecraft universe. Specifically, we develop JARVIS-1 on top of pre-trained multimodal language models, which map visual observations and textual instructions to plans. The plans will be ultimately dispatched to the goal-conditioned controllers. We outfit JARVIS-1 with a multimodal memory, which facilitates planning using both pre-trained knowledge and its actual game survival experiences. JARVIS-1 is the existing most general agent in Minecraft, capable of completing over 200 different tasks using control and observation space similar to humans. These tasks range from short-horizon tasks, e.g., "chopping trees" to long-horizon tasks, e.g., "obtaining a diamond pickaxe". JARVIS-1 performs exceptionally well in short-horizon tasks, achieving nearly perfect performance. In the classic long-term task of $\texttt{ObtainDiamondPickaxe}$, JARVIS-1 surpasses the reliability of current state-of-the-art agents by 5 times and can successfully complete longer-horizon and more challenging tasks. The project page is available at https://craftjarvis.org/JARVIS-1

Jiaming Sun, Zihao Wang, Siyu Zhang et al.

We propose a new method named OnePose for object pose estimation. Unlike existing instance-level or category-level methods, OnePose does not rely on CAD models and can handle objects in arbitrary categories without instance- or category-specific network training. OnePose draws the idea from visual localization and only requires a simple RGB video scan of the object to build a sparse SfM model of the object. Then, this model is registered to new query images with a generic feature matching network. To mitigate the slow runtime of existing visual localization methods, we propose a new graph attention network that directly matches 2D interest points in the query image with the 3D points in the SfM model, resulting in efficient and robust pose estimation. Combined with a feature-based pose tracker, OnePose is able to stably detect and track 6D poses of everyday household objects in real-time. We also collected a large-scale dataset that consists of 450 sequences of 150 objects.

Shaofei Cai, Zihao Wang, Xiaojian Ma et al. · pku

We study the problem of learning goal-conditioned policies in Minecraft, a popular, widely accessible yet challenging open-ended environment for developing human-level multi-task agents. We first identify two main challenges of learning such policies: 1) the indistinguishability of tasks from the state distribution, due to the vast scene diversity, and 2) the non-stationary nature of environment dynamics caused by partial observability. To tackle the first challenge, we propose Goal-Sensitive Backbone (GSB) for the policy to encourage the emergence of goal-relevant visual state representations. To tackle the second challenge, the policy is further fueled by an adaptive horizon prediction module that helps alleviate the learning uncertainty brought by the non-stationary dynamics. Experiments on 20 Minecraft tasks show that our method significantly outperforms the best baseline so far; in many of them, we double the performance. Our ablation and exploratory studies then explain how our approach beat the counterparts and also unveil the surprising bonus of zero-shot generalization to new scenes (biomes). We hope our agent could help shed some light on learning goal-conditioned, multi-task agents in challenging, open-ended environments like Minecraft.

Yabo Zhang, Zihao Wang, Jun Hao Liew et al.

In this work, we investigate performing semantic segmentation solely through the training on image-sentence pairs. Due to the lack of dense annotations, existing text-supervised methods can only learn to group an image into semantic regions via pixel-insensitive feedback. As a result, their grouped results are coarse and often contain small spurious regions, limiting the upper-bound performance of segmentation. On the other hand, we observe that grouped results from self-supervised models are more semantically consistent and break the bottleneck of existing methods. Motivated by this, we introduce associate self-supervised spatially-consistent grouping with text-supervised semantic segmentation. Considering the part-like grouped results, we further adapt a text-supervised model from image-level to region-level recognition with two core designs. First, we encourage fine-grained alignment with a one-way noun-to-region contrastive loss, which reduces the mismatched noun-region pairs. Second, we adopt a contextually aware masking strategy to enable simultaneous recognition of all grouped regions. Coupled with spatially-consistent grouping and region-adapted recognition, our method achieves 59.2% mIoU and 32.4% mIoU on Pascal VOC and Pascal Context benchmarks, significantly surpassing the state-of-the-art methods.

Zihao Wang, Wei Liu, Qian He et al.

Training a text-to-image generator in the general domain (e.g., Dall.e, CogView) requires huge amounts of paired text-image data, which is too expensive to collect. In this paper, we propose a self-supervised scheme named as CLIP-GEN for general text-to-image generation with the language-image priors extracted with a pre-trained CLIP model. In our approach, we only require a set of unlabeled images in the general domain to train a text-to-image generator. Specifically, given an image without text labels, we first extract the embedding of the image in the united language-vision embedding space with the image encoder of CLIP. Next, we convert the image into a sequence of discrete tokens in the VQGAN codebook space (the VQGAN model can be trained with the unlabeled image dataset in hand). Finally, we train an autoregressive transformer that maps the image tokens from its unified language-vision representation. Once trained, the transformer can generate coherent image tokens based on the text embedding extracted from the text encoder of CLIP upon an input text. Such a strategy enables us to train a strong and general text-to-image generator with large text-free image dataset such as ImageNet. Qualitative and quantitative evaluations verify that our method significantly outperforms optimization-based text-to-image methods in terms of image quality while not compromising the text-image matching. Our method can even achieve comparable performance as flagship supervised models like CogView.

Wei Zhang, Zihao Wang, Jie Fan et al.

The Gromov-Wasserstein distance is a notable extension of optimal transport. In contrast to the classic Wasserstein distance, it solves a quadratic assignment problem that minimizes the pair-wise distance distortion under the transportation of distributions and thus could apply to distributions in different spaces. These properties make Gromov-Wasserstein widely applicable to many fields, such as computer graphics and machine learning. However, the computation of the Gromov-Wasserstein distance and transport plan is expensive. The well-known Entropic Gromov-Wasserstein approach has a cubic complexity since the matrix multiplication operations need to be repeated in computing the gradient of Gromov-Wasserstein loss. This becomes a key bottleneck of the method. Currently, existing methods accelerate the computation focus on sampling and approximation, which leads to low accuracy or incomplete transport plan. In this work, we propose a novel method to accelerate accurate gradient computation by dynamic programming techniques, reducing the complexity from cubic to quadratic. In this way, the original computational bottleneck is broken and the new entropic solution can be obtained with total quadratic time, which is almost optimal complexity. Furthermore, it can be extended to some variants easily. Extensive experiments validate the efficiency and effectiveness of our method.

Zihao Wang, Chen Lin, Lu Sheng et al.

Recently, deep learning has been utilized to solve video recognition problem due to its prominent representation ability. Deep neural networks for video tasks is highly customized and the design of such networks requires domain experts and costly trial and error tests. Recent advance in network architecture search has boosted the image recognition performance in a large margin. However, automatic designing of video recognition network is less explored. In this study, we propose a practical solution, namely Practical Video Neural Architecture Search (PV-NAS).Our PV-NAS can efficiently search across tremendous large scale of architectures in a novel spatial-temporal network search space using the gradient based search methods. To avoid sticking into sub-optimal solutions, we propose a novel learning rate scheduler to encourage sufficient network diversity of the searched models. Extensive empirical evaluations show that the proposed PV-NAS achieves state-of-the-art performance with much fewer computational resources. 1) Within light-weight models, our PV-NAS-L achieves 78.7% and 62.5% Top-1 accuracy on Kinetics-400 and Something-Something V2, which are better than previous state-of-the-art methods (i.e., TSM) with a large margin (4.6% and 3.4% on each dataset, respectively), and 2) among median-weight models, our PV-NAS-M achieves the best performance (also a new record)in the Something-Something V2 dataset.

Jiayuan Gu, Wei-Chiu Ma, Sivabalan Manivasagam et al.

3D shape completion for real data is important but challenging, since partial point clouds acquired by real-world sensors are usually sparse, noisy and unaligned. Different from previous methods, we address the problem of learning 3D complete shape from unaligned and real-world partial point clouds. To this end, we propose a weakly-supervised method to estimate both 3D canonical shape and 6-DoF pose for alignment, given multiple partial observations associated with the same instance. The network jointly optimizes canonical shapes and poses with multi-view geometry constraints during training, and can infer the complete shape given a single partial point cloud. Moreover, learned pose estimation can facilitate partial point cloud registration. Experiments on both synthetic and real data show that it is feasible and promising to learn 3D shape completion through large-scale data without shape and pose supervision.

Pengcheng Wang, Zihao Wang, Zhilong Ji et al.

This paper introduces our approach to the EmotioNet Challenge 2020. We pose the AU recognition problem as a multi-task learning problem, where the non-rigid facial muscle motion (mainly the first 17 AUs) and the rigid head motion (the last 6 AUs) are modeled separately. The co-occurrence of the expression features and the head pose features are explored. We observe that different AUs converge at various speed. By choosing the optimal checkpoint for each AU, the recognition results are improved. We are able to obtain a final score of 0.746 in validation set and 0.7306 in the test set of the challenge.

Xiao Han, Zihao Wang, Enmei Tu et al.

Deep learning demands a huge amount of well-labeled data to train the network parameters. How to use the least amount of labeled data to obtain the desired classification accuracy is of great practical significance, because for many real-world applications (such as medical diagnosis), it is difficult to obtain so many labeled samples. In this paper, modify the unsupervised discriminant projection algorithm from dimension reduction and apply it as a regularization term to propose a new semi-supervised deep learning algorithm, which is able to utilize both the local and nonlocal distribution of abundant unlabeled samples to improve classification performance. Experiments show that given dozens of labeled samples, the proposed algorithm can train a deep network to attain satisfactory classification results.

Zihao Wang, Xihui Liu, Hongsheng Li et al.

Text-image cross-modal retrieval is a challenging task in the field of language and vision. Most previous approaches independently embed images and sentences into a joint embedding space and compare their similarities. However, previous approaches rarely explore the interactions between images and sentences before calculating similarities in the joint space. Intuitively, when matching between images and sentences, human beings would alternatively attend to regions in images and words in sentences, and select the most salient information considering the interaction between both modalities. In this paper, we propose Cross-modal Adaptive Message Passing (CAMP), which adaptively controls the information flow for message passing across modalities. Our approach not only takes comprehensive and fine-grained cross-modal interactions into account, but also properly handles negative pairs and irrelevant information with an adaptive gating scheme. Moreover, instead of conventional joint embedding approaches for text-image matching, we infer the matching score based on the fused features, and propose a hardest negative binary cross-entropy loss for training. Results on COCO and Flickr30k significantly surpass state-of-the-art methods, demonstrating the effectiveness of our approach.