Self-supervised denoising for massive noisy images

We propose an effective deep learning model for signal reconstruction, which requires no signal prior, no noise model calibration, and no clean samples. This model only assumes that the noise is independent of the measurement and that the true signals share the same structured information. We demonstrate its performance on a variety of real-world applications, from sub-\r{A}ngstr\"{o}m resolution atomic images to sub-arcsecond resolution astronomy images.

Self-Supervised Learning by Estimating Twin Class Distributions

We present TWIST, a novel self-supervised representation learning method by classifying large-scale unlabeled datasets in an end-to-end way. We employ a siamese network terminated by a softmax operation to produce twin class distributions of two augmented images. Without supervision, we enforce the class distributions of different augmentations to be consistent. In the meantime, we regularize the class distributions to make them sharp and diverse. Specifically, we minimize the entropy of the distribution for each sample to make the class prediction for each sample assertive and maximize the entropy of the mean distribution to make the predictions of different samples diverse. In this way, TWIST can naturally avoid the trivial solutions without specific designs such as asymmetric network, stop-gradient operation, or momentum encoder. Different from the clustering-based methods which alternate between clustering and learning, our method is a single learning process guided by a unified loss function. As a result, TWIST outperforms state-of-the-art methods on a wide range of tasks, including unsupervised classification, linear classification, semi-supervised learning, transfer learning, and some dense prediction tasks such as detection and segmentation. Codes and pre-trained models are given on: https://github.com/bytedance/TWIST

Boost Neural Networks by Checkpoints

Training multiple deep neural networks (DNNs) and averaging their outputs is a simple way to improve the predictive performance. Nevertheless, the multiplied training cost prevents this ensemble method to be practical and efficient. Several recent works attempt to save and ensemble the checkpoints of DNNs, which only requires the same computational cost as training a single network. However, these methods suffer from either marginal accuracy improvements due to the low diversity of checkpoints or high risk of divergence due to the cyclical learning rates they adopted. In this paper, we propose a novel method to ensemble the checkpoints, where a boosting scheme is utilized to accelerate model convergence and maximize the checkpoint diversity. We theoretically prove that it converges by reducing exponential loss. The empirical evaluation also indicates our proposed ensemble outperforms single model and existing ensembles in terms of accuracy and efficiency. With the same training budget, our method achieves 4.16% lower error on Cifar-100 and 6.96% on Tiny-ImageNet with ResNet-110 architecture. Moreover, the adaptive sample weights in our method make it an effective solution to address the imbalanced class distribution. In the experiments, it yields up to 5.02% higher accuracy over single EfficientNet-B0 on the imbalanced datasets.

MC$^2$-SF: Slow-Fast Learning for Mobile-Cloud Collaborative Recommendation

With the hardware development of mobile devices, it is possible to build the recommendation models on the mobile side to utilize the fine-grained features and the real-time feedbacks. Compared to the straightforward mobile-based modeling appended to the cloud-based modeling, we propose a Slow-Fast learning mechanism to make the Mobile-Cloud Collaborative recommendation (MC$^2$-SF) mutual benefit. Specially, in our MC$^2$-SF, the cloud-based model and the mobile-based model are respectively treated as the slow component and the fast component, according to their interaction frequency in real-world scenarios. During training and serving, they will communicate the prior/privileged knowledge to each other to help better capture the user interests about the candidates, resembling the role of System I and System II in the human cognition. We conduct the extensive experiments on three benchmark datasets and demonstrate the proposed MC$^2$-SF outperforms several state-of-the-art methods.

Less is More: Learning from Synthetic Data with Fine-grained Attributes for Person Re-Identification

Person re-identification (re-ID) plays an important role in applications such as public security and video surveillance. Recently, learning from synthetic data, which benefits from the popularity of synthetic data engine, has attracted attention from both academia and the public eye. However, existing synthetic datasets are limited in quantity, diversity and realisticity, and cannot be efficiently used for generalizable re-ID problem. To address this challenge, we construct and label a large-scale synthetic person dataset named FineGPR with fine-grained attribute distribution. Moreover, aiming to fully exploit the potential of FineGPR and promote the efficient training from millions of synthetic data, we propose an attribute analysis pipeline AOST to learn attribute distribution in target domain, then apply style transfer network to eliminate the gap between synthetic and real-world data and thus is freely deployed to new scenarios. Experiments conducted on benchmarks demonstrate that FineGPR with AOST outperforms (or is on par with) existing real and synthetic datasets, which suggests its feasibility for re-ID and proves the proverbial less-is-more principle. We hope this fine-grained dataset could advance research towards re-ID in real scenarios.

Dynamic RAT Selection and Transceiver Optimization for Mobile Edge Computing Over Multi-RAT Heterogeneous Networks

Mobile edge computing (MEC) integrated with multiple radio access technologies (RATs) is a promising technique for satisfying the growing low-latency computation demand of emerging intelligent internet of things (IoT) applications. Under the distributed MapReduce framework, this paper investigates the joint RAT selection and transceiver design for over-the-air (OTA) aggregation of intermediate values (IVAs) in wireless multiuser MEC systems, while taking into account the energy budget constraint for the local computing and IVA transmission per wireless device (WD). We aim to minimize the weighted sum of the computation mean squared error (MSE) of the aggregated IVA at the RAT receivers, the WDs' IVA transmission cost, and the associated transmission time delay, which is a mixed-integer and non-convex problem. Based on the Lagrange duality method and primal decomposition, we develop a low-complexity algorithm by solving the WDs' RAT selection problem, the WDs' transmit coefficients optimization problem, and the aggregation beamforming problem. Extensive numerical results are provided to demonstrate the effectiveness and merit of our proposed algorithm as compared with other existing schemes.

YOLOX: Exceeding YOLO Series in 2021

In this report, we present some experienced improvements to YOLO series, forming a new high-performance detector -- YOLOX. We switch the YOLO detector to an anchor-free manner and conduct other advanced detection techniques, i.e., a decoupled head and the leading label assignment strategy SimOTA to achieve state-of-the-art results across a large scale range of models: For YOLO-Nano with only 0.91M parameters and 1.08G FLOPs, we get 25.3% AP on COCO, surpassing NanoDet by 1.8% AP; for YOLOv3, one of the most widely used detectors in industry, we boost it to 47.3% AP on COCO, outperforming the current best practice by 3.0% AP; for YOLOX-L with roughly the same amount of parameters as YOLOv4-CSP, YOLOv5-L, we achieve 50.0% AP on COCO at a speed of 68.9 FPS on Tesla V100, exceeding YOLOv5-L by 1.8% AP. Further, we won the 1st Place on Streaming Perception Challenge (Workshop on Autonomous Driving at CVPR 2021) using a single YOLOX-L model. We hope this report can provide useful experience for developers and researchers in practical scenes, and we also provide deploy versions with ONNX, TensorRT, NCNN, and Openvino supported. Source code is at https://github.com/Megvii-BaseDetection/YOLOX.

Computation Rate Maximization for Multiuser Mobile Edge Computing Systems With Dynamic Energy Arrivals

This paper considers an energy harvesting (EH) based multiuser mobile edge computing (MEC) system, where each user utilizes the harvested energy from renewable energy sources to execute its computation tasks via computation offloading and local computing. Towards maximizing the system's weighted computation rate (i.e., the number of weighted users' computing bits within a finite time horizon) subject to the users' energy causality constraints due to dynamic energy arrivals, the decision for joint computation offloading and local computing over time is optimized {\em over time}. Assuming that the profile of channel state information and dynamic task arrivals at the users is known in advance, the weighted computation rate maximization problem becomes a convex optimization problem. Building on the Lagrange duality method, the well-structured optimal solution is analytically obtained. Both the users' local computing and offloading rates are shown to have a monotonically increasing structure. Numerical results show that the proposed design scheme can achieve a significant performance gain over the alternative benchmark schemes.