Unifying Visual Perception by Dispersible Points Learning

We present a conceptually simple, flexible, and universal visual perception head for variant visual tasks, e.g., classification, object detection, instance segmentation and pose estimation, and different frameworks, such as one-stage or two-stage pipelines. Our approach effectively identifies an object in an image while simultaneously generating a high-quality bounding box or contour-based segmentation mask or set of keypoints. The method, called UniHead, views different visual perception tasks as the dispersible points learning via the transformer encoder architecture. Given a fixed spatial coordinate, UniHead adaptively scatters it to different spatial points and reasons about their relations by transformer encoder. It directly outputs the final set of predictions in the form of multiple points, allowing us to perform different visual tasks in different frameworks with the same head design. We show extensive evaluations on ImageNet classification and all three tracks of the COCO suite of challenges, including object detection, instance segmentation and pose estimation. Without bells and whistles, UniHead can unify these visual tasks via a single visual head design and achieve comparable performance compared to expert models developed for each task.We hope our simple and universal UniHead will serve as a solid baseline and help promote universal visual perception research. Code and models are available at https://github.com/Sense-X/UniHead.

Self-slimmed Vision Transformer

Vision transformers (ViTs) have become the popular structures and outperformed convolutional neural networks (CNNs) on various vision tasks. However, such powerful transformers bring a huge computation burden. And the essential barrier behind this is the exhausting token-to-token comparison. To alleviate this, we delve deeply into the model properties of ViT and observe that ViTs exhibit sparse attention with high token similarity. This intuitively introduces us a feasible structure-agnostic dimension, token number, to reduce the computational cost. Based on this exploration, we propose a generic self-slimmed learning approach for vanilla ViTs, namely SiT. Specifically, we first design a novel Token Slimming Module (TSM), which can boost the inference efficiency of ViTs by dynamic token aggregation. Different from the token hard dropping, our TSM softly integrates redundant tokens into fewer informative ones, which can dynamically zoom visual attention without cutting off discriminative token relations in the images. Furthermore, we introduce a concise Dense Knowledge Distillation (DKD) framework, which densely transfers unorganized token information in a flexible auto-encoder manner. Due to the similar structure between teacher and student, our framework can effectively leverage structure knowledge for better convergence. Finally, we conduct extensive experiments to evaluate our SiT. It demonstrates that our method can speed up ViTs by 1.7x with negligible accuracy drop, and even speed up ViTs by 3.6x while maintaining 97% of their performance. Surprisingly, by simply arming LV-ViT with our SiT, we achieve new state-of-the-art performance on ImageNet, surpassing all the CNNs and ViTs in the recent literature.

RCNet: Reverse Feature Pyramid and Cross-scale Shift Network for Object Detection

Feature pyramid networks (FPN) are widely exploited for multi-scale feature fusion in existing advanced object detection frameworks. Numerous previous works have developed various structures for bidirectional feature fusion, all of which are shown to improve the detection performance effectively. We observe that these complicated network structures require feature pyramids to be stacked in a fixed order, which introduces longer pipelines and reduces the inference speed. Moreover, semantics from non-adjacent levels are diluted in the feature pyramid since only features at adjacent pyramid levels are merged by the local fusion operation in a sequence manner. To address these issues, we propose a novel architecture named RCNet, which consists of Reverse Feature Pyramid (RevFP) and Cross-scale Shift Network (CSN). RevFP utilizes local bidirectional feature fusion to simplify the bidirectional pyramid inference pipeline. CSN directly propagates representations to both adjacent and non-adjacent levels to enable multi-scale features more correlative. Extensive experiments on the MS COCO dataset demonstrate RCNet can consistently bring significant improvements over both one-stage and two-stage detectors with subtle extra computational overhead. In particular, RetinaNet is boosted to 40.2 AP, which is 3.7 points higher than baseline, by replacing FPN with our proposed model. On COCO test-dev, RCNet can achieve very competitive performance with a single-model single-scale 50.5 AP. Codes will be made available.

RaidaR: A Rich Annotated Image Dataset of Rainy Street Scenes

We introduce RaidaR, a rich annotated image dataset of rainy street scenes, to support autonomous driving research. The new dataset contains the largest number of rainy images (58,542) to date, 5,000 of which provide semantic segmentations and 3,658 provide object instance segmentations. The RaidaR images cover a wide range of realistic rain-induced artifacts, including fog, droplets, and road reflections, which can effectively augment existing street scene datasets to improve data-driven machine perception during rainy weather. To facilitate efficient annotation of a large volume of images, we develop a semi-automatic scheme combining manual segmentation and an automated processing akin to cross validation, resulting in 10-20 fold reduction on annotation time. We demonstrate the utility of our new dataset by showing how data augmentation with RaidaR can elevate the accuracy of existing segmentation algorithms. We also present a novel unpaired image-to-image translation algorithm for adding/removing rain artifacts, which directly benefits from RaidaR.

KPNet: Towards Minimal Face Detector

The small receptive field and capacity of minimal neural networks limit their performance when using them to be the backbone of detectors. In this work, we find that the appearance feature of a generic face is discriminative enough for a tiny and shallow neural network to verify from the background. And the essential barriers behind us are 1) the vague definition of the face bounding box and 2) tricky design of anchor-boxes or receptive field. Unlike most top-down methods for joint face detection and alignment, the proposed KPNet detects small facial keypoints instead of the whole face by in a bottom-up manner. It first predicts the facial landmarks from a low-resolution image via the well-designed fine-grained scale approximation and scale adaptive soft-argmax operator. Finally, the precise face bounding boxes, no matter how we define it, can be inferred from the keypoints. Without any complex head architecture or meticulous network designing, the KPNet achieves state-of-the-art accuracy on generic face detection and alignment benchmarks with only $\sim1M$ parameters, which runs at 1000fps on GPU and is easy to perform real-time on most modern front-end chips.

Rethinking Loss Design for Large-scale 3D Shape Retrieval

Learning discriminative shape representations is a crucial issue for large-scale 3D shape retrieval. In this paper, we propose the Collaborative Inner Product Loss (CIP Loss) to obtain ideal shape embedding that discriminative among different categories and clustered within the same class. Utilizing simple inner product operation, CIP loss explicitly enforces the features of the same class to be clustered in a linear subspace, while inter-class subspaces are constrained to be at least orthogonal. Compared to previous metric loss functions, CIP loss could provide more clear geometric interpretation for the embedding than Euclidean margin, and is easy to implement without normalization operation referring to cosine margin. Moreover, our proposed loss term can combine with other commonly used loss functions and can be easily plugged into existing off-the-shelf architectures. Extensive experiments conducted on the two public 3D object retrieval datasets, ModelNet and ShapeNetCore 55, demonstrate the effectiveness of our proposal, and our method has achieved state-of-the-art results on both datasets.

Angular Triplet-Center Loss for Multi-view 3D Shape Retrieval

How to obtain the desirable representation of a 3D shape, which is discriminative across categories and polymerized within classes, is a significant challenge in 3D shape retrieval. Most existing 3D shape retrieval methods focus on capturing strong discriminative shape representation with softmax loss for the classification task, while the shape feature learning with metric loss is neglected for 3D shape retrieval. In this paper, we address this problem based on the intuition that the cosine distance of shape embeddings should be close enough within the same class and far away across categories. Since most of 3D shape retrieval tasks use cosine distance of shape features for measuring shape similarity, we propose a novel metric loss named angular triplet-center loss, which directly optimizes the cosine distances between the features. It inherits the triplet-center loss property to achieve larger inter-class distance and smaller intra-class distance simultaneously. Unlike previous metric loss utilized in 3D shape retrieval methods, where Euclidean distance is adopted and the margin design is difficult, the proposed method is more convenient to train feature embeddings and more suitable for 3D shape retrieval. Moreover, the angle margin is adopted to replace the cosine margin in order to provide more explicit discriminative constraints on an embedding space. Extensive experimental results on two popular 3D object retrieval benchmarks, ModelNet40 and ShapeNetCore 55, demonstrate the effectiveness of our proposed loss, and our method has achieved state-of-the-art results on various 3D shape datasets.

Learning Discriminative 3D Shape Representations by View Discerning Networks

In view-based 3D shape recognition, extracting discriminative visual representation of 3D shapes from projected images is considered the core problem. Projections with low discriminative ability can adversely influence the final 3D shape representation. Especially under the real situations with background clutter and object occlusion, the adverse effect is even more severe. To resolve this problem, we propose a novel deep neural network, View Discerning Network, which learns to judge the quality of views and adjust their contributions to the representation of shapes. In this network, a Score Generation Unit is devised to evaluate the quality of each projected image with score vectors. These score vectors are used to weight the image features and the weighted features perform much better than original features in 3D shape recognition task. In particular, we introduce two structures of Score Generation Unit, Channel-wise Score Unit and Part-wise Score Unit, to assess the quality of feature maps from different perspectives. Our network aggregates features and scores in an end-to-end framework, so that final shape descriptors are directly obtained from its output. Our experiments on ModelNet and ShapeNet Core55 show that View Discerning Network outperforms the state-of-the-arts in terms of the retrieval task, with excellent robustness against background clutter and object occlusion.

Beyond Trade-off: Accelerate FCN-based Face Detector with Higher Accuracy

Fully convolutional neural network (FCN) has been dominating the game of face detection task for a few years with its congenital capability of sliding-window-searching with shared kernels, which boiled down all the redundant calculation, and most recent state-of-the-art methods such as Faster-RCNN, SSD, YOLO and FPN use FCN as their backbone. So here comes one question: Can we find a universal strategy to further accelerate FCN with higher accuracy, so could accelerate all the recent FCN-based methods? To analyze this, we decompose the face searching space into two orthogonal directions, `scale' and `spatial'. Only a few coordinates in the space expanded by the two base vectors indicate foreground. So if FCN could ignore most of the other points, the searching space and false alarm should be significantly boiled down. Based on this philosophy, a novel method named scale estimation and spatial attention proposal ($S^2AP$) is proposed to pay attention to some specific scales and valid locations in the image pyramid. Furthermore, we adopt a masked-convolution operation based on the attention result to accelerate FCN calculation. Experiments show that FCN-based method RPN can be accelerated by about $4\times$ with the help of $S^2AP$ and masked-FCN and at the same time it can also achieve the state-of-the-art on FDDB, AFW and MALF face detection benchmarks as well.

Region-based Quality Estimation Network for Large-scale Person Re-identification

One of the major restrictions on the performance of video-based person re-id is partial noise caused by occlusion, blur and illumination. Since different spatial regions of a single frame have various quality, and the quality of the same region also varies across frames in a tracklet, a good way to address the problem is to effectively aggregate complementary information from all frames in a sequence, using better regions from other frames to compensate the influence of an image region with poor quality. To achieve this, we propose a novel Region-based Quality Estimation Network (RQEN), in which an ingenious training mechanism enables the effective learning to extract the complementary region-based information between different frames. Compared with other feature extraction methods, we achieved comparable results of 92.4%, 76.1% and 77.83% on the PRID 2011, iLIDS-VID and MARS, respectively. In addition, to alleviate the lack of clean large-scale person re-id datasets for the community, this paper also contributes a new high-quality dataset, named "Labeled Pedestrian in the Wild (LPW)" which contains 7,694 tracklets with over 590,000 images. Despite its relatively large scale, the annotations also possess high cleanliness. Moreover, it's more challenging in the following aspects: the age of characters varies from childhood to elderhood; the postures of people are diverse, including running and cycling in addition to the normal walking state.