SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers

We present SegFormer, a simple, efficient yet powerful semantic segmentation framework which unifies Transformers with lightweight multilayer perception (MLP) decoders. SegFormer has two appealing features: 1) SegFormer comprises a novel hierarchically structured Transformer encoder which outputs multiscale features. It does not need positional encoding, thereby avoiding the interpolation of positional codes which leads to decreased performance when the testing resolution differs from training. 2) SegFormer avoids complex decoders. The proposed MLP decoder aggregates information from different layers, and thus combining both local attention and global attention to render powerful representations. We show that this simple and lightweight design is the key to efficient segmentation on Transformers. We scale our approach up to obtain a series of models from SegFormer-B0 to SegFormer-B5, reaching significantly better performance and efficiency than previous counterparts. For example, SegFormer-B4 achieves 50.3% mIoU on ADE20K with 64M parameters, being 5x smaller and 2.2% better than the previous best method. Our best model, SegFormer-B5, achieves 84.0% mIoU on Cityscapes validation set and shows excellent zero-shot robustness on Cityscapes-C. Code will be released at: github.com/NVlabs/SegFormer.

Reconfigurable Intelligent Surface-Aided Wireless Power Transfer Systems: Analysis and Implementation

Reconfigurable intelligent surface (RIS) is a promising technology for RF wireless power transfer (WPT) as it is capable of beamforming and beam focusing without using active and power-hungry components. In this paper, we propose a multi-tile RIS beam scanning (MTBS) algorithm for powering up internet-of-things (IoT) devices. Considering the hardware limitations of the IoT devices, the proposed algorithm requires only power information to enable the beam focusing capability of the RIS. Specifically, we first divide the RIS into smaller RIS tiles. Then, all RIS tiles and the phased array transmitter are iteratively scanned and optimized to maximize the receive power. We elaborately analyze the proposed algorithm and build a simulator to verify it. Furthermore, we have built a real-life testbed of RIS-aided WPT systems to validate the algorithm. The experimental results show that the proposed MTBS algorithm can properly control the transmission phase of the transmitter and the reflection phase of the RIS to focus the power at the receiver. Consequently, after executing the algorithm, about 20 dB improvement of the receive power is achieved compared to the case that all unit cells of the RIS are in OFF state. By experiments, we confirm that the RIS with the MTBS algorithm can greatly enhance the power transfer efficiency.

A Self-Guided Approach for Navigation in a Minimalistic Foraging Robotic Swarm

We present a biologically inspired design for swarm foraging based on ant's pheromone deployment, where the swarm is assumed to have very restricted capabilities. The robots do not require global or relative position measurements and the swarm is fully decentralized and needs no infrastructure in place. Additionally, the system only requires one-hop communication over the robot network, we do not make any assumptions about the connectivity of the communication graph and the transmission of information and computation is scalable versus the number of agents. This is done by letting the agents in the swarm act as foragers or as guiding agents (beacons). We present experimental results computed for a swarm of Elisa-3 robots on a simulator, and show how the swarm self-organizes to solve a foraging problem over an unknown environment, converging to trajectories around the shortest path. At last, we discuss the limitations of such a system and propose how the foraging efficiency can be increased.

LPF: A Language-Prior Feedback Objective Function for De-biased Visual Question Answering

Most existing Visual Question Answering (VQA) systems tend to overly rely on language bias and hence fail to reason from the visual clue. To address this issue, we propose a novel Language-Prior Feedback (LPF) objective function, to re-balance the proportion of each answer's loss value in the total VQA loss. The LPF firstly calculates a modulating factor to determine the language bias using a question-only branch. Then, the LPF assigns a self-adaptive weight to each training sample in the training process. With this reweighting mechanism, the LPF ensures that the total VQA loss can be reshaped to a more balanced form. By this means, the samples that require certain visual information to predict will be efficiently used during training. Our method is simple to implement, model-agnostic, and end-to-end trainable. We conduct extensive experiments and the results show that the LPF (1) brings a significant improvement over various VQA models, (2) achieves competitive performance on the bias-sensitive VQA-CP v2 benchmark.

Document Collection Visual Question Answering

Current tasks and methods in Document Understanding aims to process documents as single elements. However, documents are usually organized in collections (historical records, purchase invoices), that provide context useful for their interpretation. To address this problem, we introduce Document Collection Visual Question Answering (DocCVQA) a new dataset and related task, where questions are posed over a whole collection of document images and the goal is not only to provide the answer to the given question, but also to retrieve the set of documents that contain the information needed to infer the answer. Along with the dataset we propose a new evaluation metric and baselines which provide further insights to the new dataset and task.

Joint Client Scheduling and Resource Allocation under Channel Uncertainty in Federated Learning

The performance of federated learning (FL) over wireless networks depend on the reliability of the client-server connectivity and clients' local computation capabilities. In this article we investigate the problem of client scheduling and resource block (RB) allocation to enhance the performance of model training using FL, over a pre-defined training duration under imperfect channel state information (CSI) and limited local computing resources. First, we analytically derive the gap between the training losses of FL with clients scheduling and a centralized training method for a given training duration. Then, we formulate the gap of the training loss minimization over client scheduling and RB allocation as a stochastic optimization problem and solve it using Lyapunov optimization. A Gaussian process regression-based channel prediction method is leveraged to learn and track the wireless channel, in which, the clients' CSI predictions and computing power are incorporated into the scheduling decision. Using an extensive set of simulations, we validate the robustness of the proposed method under both perfect and imperfect CSI over an array of diverse data distributions. Results show that the proposed method reduces the gap of the training accuracy loss by up to 40.7% compared to state-of-theart client scheduling and RB allocation methods.

The Gaze and Mouse Signal as additional Source for User Fingerprints in Browser Applications

In this work we inspect different data sources for browser fingerprints. We show which disadvantages and limitations browser statistics have and how this can be avoided with other data sources. Since human visual behavior is a rich source of information and also contains person specific information, it is a valuable source for browser fingerprints. However, human gaze acquisition in the browser also has disadvantages, such as inaccuracies via webcam and the restriction that the user must first allow access to the camera. However, it is also known that the mouse movements and the human gaze correlate and therefore, the mouse movements can be used instead of the gaze signal. In our evaluation we show the influence of all possible combinations of the three information sources for user recognition and describe our simple approach in detail. The data and the Matlab code can be downloaded here https://atreus.informatik.uni-tuebingen.de/seafile/d/8e2ab8c3fdd444e1a135/?p=%2FThe%20Gaze%20and%20Mouse%20Signal%20as%20additional%20Source%20...&mode=list

Mini-batch graphs for robust image classification

Current deep learning models for classification tasks in computer vision are trained using mini-batches. In the present article, we take advantage of the relationships between samples in a mini-batch, using graph neural networks to aggregate information from similar images. This helps mitigate the adverse effects of alterations to the input images on classification performance. Diverse experiments on image-based object and scene classification show that this approach not only improves a classifier's performance but also increases its robustness to image perturbations and adversarial attacks. Further, we also show that mini-batch graph neural networks can help to alleviate the problem of mode collapse in Generative Adversarial Networks.

Automatic Classification of Games using Support Vector Machine

Game developers benefit from availability of custom game genres when doing game market analysis. This information can help them to spot opportunities in market and make them more successful in planning a new game. In this paper we find good classifier for predicting category of a game. Prediction is based on description and title of a game. We use 2443 iOS App Store games as data set to generate a document-term matrix. To reduce the curse of dimensionality we use Latent Semantic Indexing, which, reduces the term dimension to approximately 1/9. Support Vector Machine supervised learning model is fit to pre-processed data. Model parameters are optimized using grid search and 20-fold cross validation. Best model yields to 77% mean accuracy or roughly 70% accuracy with 95% confidence. Developed classifier has been used in-house to assist games market research.

Random Hash Code Generation for Cancelable Fingerprint Templates using Vector Permutation and Shift-order Process

Cancelable biometric techniques have been used to prevent the compromise of biometric data by generating and using their corresponding cancelable templates for user authentication. However, the non-invertible distance preserving transformation methods employed in various schemes are often vulnerable to information leakage since matching is performed in the transformed domain. In this paper, we propose a non-invertible distance preserving scheme based on vector permutation and shift-order process. First, the dimension of feature vectors is reduced using kernelized principle component analysis (KPCA) prior to randomly permuting the extracted vector features. A shift-order process is then applied to the generated features in order to achieve non-invertibility and combat similarity-based attacks. The generated hash codes are resilient to different security and privacy attacks whilst fulfilling the major revocability and unlinkability requirements. Experimental evaluation conducted on 6 datasets of FVC2002 and FVC2004 reveals a high-performance accuracy of the proposed scheme better than other existing state-of-the-art schemes.