Developing a unified multi-task foundation model has become a critical challenge in computer vision research. In the current field of 3D computer vision, most datasets only focus on single task, which complicates the concurrent training requirements of various downstream tasks. In this paper, we introduce VEnvision3D, a large 3D synthetic perception dataset for multi-task learning, including depth completion, segmentation, upsampling, place recognition, and 3D reconstruction. Since the data for each task is collected in the same environmental domain, sub-tasks are inherently aligned in terms of the utilized data. Therefore, such a unique attribute can assist in exploring the potential for the multi-task model and even the foundation model without separate training methods. Meanwhile, capitalizing on the advantage of virtual environments being freely editable, we implement some novel settings such as simulating temporal changes in the environment and sampling point clouds on model surfaces. These characteristics enable us to present several new benchmarks. We also perform extensive studies on multi-task end-to-end models, revealing new observations, challenges, and opportunities for future research. Our dataset and code will be open-sourced upon acceptance.
We propose SparseDC, a model for Depth Completion of Sparse and non-uniform depth inputs. Unlike previous methods focusing on completing fixed distributions on benchmark datasets (e.g., NYU with 500 points, KITTI with 64 lines), SparseDC is specifically designed to handle depth maps with poor quality in real usage. The key contributions of SparseDC are two-fold. First, we design a simple strategy, called SFFM, to improve the robustness under sparse input by explicitly filling the unstable depth features with stable image features. Second, we propose a two-branch feature embedder to predict both the precise local geometry of regions with available depth values and accurate structures in regions with no depth. The key of the embedder is an uncertainty-based fusion module called UFFM to balance the local and long-term information extracted by CNNs and ViTs. Extensive indoor and outdoor experiments demonstrate the robustness of our framework when facing sparse and non-uniform input depths. The pre-trained model and code are available at https://github.com/WHU-USI3DV/SparseDC.
Point cloud upsampling is to densify a sparse point set acquired from 3D sensors, providing a denser representation for underlying surface. However, existing methods perform upsampling on a single patch, ignoring the coherence and relation of the entire surface, thus limiting the upsampled capability. Also, they mainly focus on a clean input, thus the performance is severely compromised when handling scenarios with extra noises. In this paper, we present a novel method for more effective point cloud upsampling, achieving a more robust and improved performance. To this end, we incorporate two thorough considerations. i) Instead of upsampling each small patch independently as previous works, we take adjacent patches as input and introduce a Patch Correlation Unit to explore the shape correspondence between them for effective upsampling. ii)We propose a Position Correction Unit to mitigate the effects of outliers and noisy points. It contains a distance-aware encoder to dynamically adjust the generated points to be close to the underlying surface. Extensive experiments demonstrate that our proposed method surpasses previous upsampling methods on both clean and noisy inputs.
We propose a dynamic boosted ensemble learning method based on random forest (DBRF), a novel ensemble algorithm that incorporates the notion of hard example mining into Random Forest (RF) and thus combines the high accuracy of Boosting algorithm with the strong generalization of Bagging algorithm. Specifically, we propose to measure the quality of each leaf node of every decision tree in the random forest to determine hard examples. By iteratively training and then removing easy examples from training data, we evolve the random forest to focus on hard examples dynamically so as to learn decision boundaries better. Data can be cascaded through these random forests learned in each iteration in sequence to generate predictions, thus making RF deep. We also propose to use evolution mechanism and smart iteration mechanism to improve the performance of the model. DBRF outperforms RF on three UCI datasets and achieved state-of-the-art results compared to other deep models. Moreover, we show that DBRF is also a new way of sampling and can be very useful when learning from imbalanced data.