The objective of this work is to explore how to effectively and efficiently adapt pre-trained foundation models to various downstream tasks of image semantic segmentation. Conventional methods usually fine-tuned the whole networks for each specific dataset and it was burdensome to store the massive parameters of these networks. A few recent works attempted to insert some trainable parameters into the frozen network to learn visual prompts for efficient tuning. However, these works significantly modified the original structure of standard modules, making them inoperable on many existing high-speed inference devices, where standard modules and their parameters have been embedded. To facilitate prompt-based semantic segmentation, we propose a novel Inter-Stage Prompt-Matched Framework, which maintains the original structure of the foundation model while generating visual prompts adaptively for task-oriented tuning. Specifically, the pre-trained model is first divided into multiple stages, and their parameters are frozen and shared for all semantic segmentation tasks. A lightweight module termed Semantic-aware Prompt Matcher is then introduced to hierarchically interpolate between two stages to learn reasonable prompts for each specific task under the guidance of interim semantic maps. In this way, we can better stimulate the pre-trained knowledge of the frozen model to learn semantic concepts effectively on downstream datasets. Extensive experiments conducted on five benchmarks show that the proposed method can achieve a promising trade-off between parameter efficiency and performance effectiveness.
In computer vision, fine-tuning is the de-facto approach to leverage pre-trained vision models to perform downstream tasks. However, deploying it in practice is quite challenging, due to adopting parameter inefficient global update and heavily relying on high-quality downstream data. Recently, prompt-based learning, which adds a task-relevant prompt to adapt the downstream tasks to pre-trained models, has drastically boosted the performance of many natural language downstream tasks. In this work, we extend this notable transfer ability benefited from prompt into vision models as an alternative to fine-tuning. To this end, we propose parameter-efficient Prompt tuning (Pro-tuning) to adapt frozen vision models to various downstream vision tasks. The key to Pro-tuning is prompt-based tuning, i.e., learning task-specific vision prompts for downstream input images with the pre-trained model frozen. By only training a few additional parameters, it can work on diverse CNN-based and Transformer-based architectures. Extensive experiments evidence that Pro-tuning outperforms fine-tuning in a broad range of vision tasks and scenarios, including image classification (generic objects, class imbalance, image corruption, adversarial robustness, and out-of-distribution generalization), and dense prediction tasks such as object detection and semantic segmentation.
Fine-grained object retrieval aims to learn discriminative representation to retrieve visually similar objects. However, existing top-performing works usually impose pairwise similarities on the semantic embedding spaces to continually fine-tune the entire model in limited-data regimes, thus resulting in easily converging to suboptimal solutions. In this paper, we develop Fine-grained Retrieval Prompt Tuning (FRPT), which steers a frozen pre-trained model to perform the fine-grained retrieval task from the perspectives of sample prompt and feature adaptation. Specifically, FRPT only needs to learn fewer parameters in the prompt and adaptation instead of fine-tuning the entire model, thus solving the convergence to suboptimal solutions caused by fine-tuning the entire model. Technically, as sample prompts, a structure perturbation prompt (SPP) is introduced to zoom and even exaggerate some pixels contributing to category prediction via a content-aware inhomogeneous sampling operation. In this way, SPP can make the fine-grained retrieval task aided by the perturbation prompts close to the solved task during the original pre-training. Besides, a category-specific awareness head is proposed and regarded as feature adaptation, which removes the species discrepancies in the features extracted by the pre-trained model using instance normalization, and thus makes the optimized features only include the discrepancies among subcategories. Extensive experiments demonstrate that our FRPT with fewer learnable parameters achieves the state-of-the-art performance on three widely-used fine-grained datasets.
In the past few years, the emergence of vision-language pre-training (VLP) has brought cross-modal retrieval to a new era. However, due to the latency and computation demand, it is commonly challenging to apply VLP in a real-time online retrieval system. To alleviate the defect, this paper proposes a \textbf{Hi}erarchical \textbf{V}ision-\textbf{}Language \textbf{P}re-Training (\textbf{HiVLP}) for fast Image-Text Retrieval (ITR). Specifically, we design a novel hierarchical retrieval objective, which uses the representation of different dimensions for coarse-to-fine ITR, i.e., using low-dimensional representation for large-scale coarse retrieval and high-dimensional representation for small-scale fine retrieval. We evaluate our proposed HiVLP on two popular image-text retrieval benchmarks, i.e., Flickr30k and COCO. Extensive experiments demonstrate that our HiVLP not only has fast inference speed but also can be easily scaled to large-scale ITR scenarios. The detailed results show that HiVLP is $1,427$$\sim$$120,649\times$ faster than the fusion-based model UNITER and 2$\sim$5 faster than the fastest embedding-based model LightingDot in different candidate scenarios. It also achieves about +4.9 AR on COCO and +3.8 AR on Flickr30K than LightingDot and achieves comparable performance with the state-of-the-art (SOTA) fusion-based model METER.
Exploiting convolutional neural networks for point cloud processing is quite challenging, due to the inherent irregular distribution and discrete shape representation of point clouds. To address these problems, many handcrafted convolution variants have sprung up in recent years. Though with elaborate design, these variants could be far from optimal in sufficiently capturing diverse shapes formed by discrete points. In this paper, we propose PointSeaConv, i.e., a novel differential convolution search paradigm on point clouds. It can work in a purely data-driven manner and thus is capable of auto-creating a group of suitable convolutions for geometric shape modeling. We also propose a joint optimization framework for simultaneous search of internal convolution and external architecture, and introduce epsilon-greedy algorithm to alleviate the effect of discretization error. As a result, PointSeaNet, a deep network that is sufficient to capture geometric shapes at both convolution level and architecture level, can be searched out for point cloud processing. Extensive experiments strongly evidence that our proposed PointSeaNet surpasses current handcrafted deep models on challenging benchmarks across multiple tasks with remarkable margins.
This is an opinion paper. We hope to deliver a key message that current visual recognition systems are far from complete, i.e., recognizing everything that human can recognize, yet it is very unlikely that the gap can be bridged by continuously increasing human annotations. Based on the observation, we advocate for a new type of pre-training task named learning-by-compression. The computational models (e.g., a deep network) are optimized to represent the visual data using compact features, and the features preserve the ability to recover the original data. Semantic annotations, when available, play the role of weak supervision. An important yet challenging issue is the evaluation of image recovery, where we suggest some design principles and future research directions. We hope our proposal can inspire the community to pursue the compression-recovery tradeoff rather than the accuracy-complexity tradeoff.
Recent years have witnessed significant progress in 3D hand mesh recovery. Nevertheless, because of the intrinsic 2D-to-3D ambiguity, recovering camera-space 3D information from a single RGB image remains challenging. To tackle this problem, we divide camera-space mesh recovery into two sub-tasks, i.e., root-relative mesh recovery and root recovery. First, joint landmarks and silhouette are extracted from a single input image to provide 2D cues for the 3D tasks. In the root-relative mesh recovery task, we exploit semantic relations among joints to generate a 3D mesh from the extracted 2D cues. Such generated 3D mesh coordinates are expressed relative to a root position, i.e., wrist of the hand. In the root recovery task, the root position is registered to the camera space by aligning the generated 3D mesh back to 2D cues, thereby completing cameraspace 3D mesh recovery. Our pipeline is novel in that (1) it explicitly makes use of known semantic relations among joints and (2) it exploits 1D projections of the silhouette and mesh to achieve robust registration. Extensive experiments on popular datasets such as FreiHAND, RHD, and Human3.6M demonstrate that our approach achieves stateof-the-art performance on both root-relative mesh recovery and root recovery. Our code is publicly available at https://github.com/SeanChenxy/HandMesh.