Multi-Modal 3D Object Detection in Autonomous Driving: a Survey

In the past few years, we have witnessed rapid development of autonomous driving. However, achieving full autonomy remains a daunting task due to the complex and dynamic driving environment. As a result, self-driving cars are equipped with a suite of sensors to conduct robust and accurate environment perception. As the number and type of sensors keep increasing, combining them for better perception is becoming a natural trend. So far, there has been no indepth review that focuses on multi-sensor fusion based perception. To bridge this gap and motivate future research, this survey devotes to review recent fusion-based 3D detection deep learning models that leverage multiple sensor data sources, especially cameras and LiDARs. In this survey, we first introduce the background of popular sensors for autonomous cars, including their common data representations as well as object detection networks developed for each type of sensor data. Next, we discuss some popular datasets for multi-modal 3D object detection, with a special focus on the sensor data included in each dataset. Then we present in-depth reviews of recent multi-modal 3D detection networks by considering the following three aspects of the fusion: fusion location, fusion data representation, and fusion granularity. After a detailed review, we discuss open challenges and point out possible solutions. We hope that our detailed review can help researchers to embark investigations in the area of multi-modal 3D object detection.

WaveGrad 2: Iterative Refinement for Text-to-Speech Synthesis

This paper introduces WaveGrad 2, a non-autoregressive generative model for text-to-speech synthesis. WaveGrad 2 is trained to estimate the gradient of the log conditional density of the waveform given a phoneme sequence. The model takes an input phoneme sequence, and through an iterative refinement process, generates an audio waveform. This contrasts to the original WaveGrad vocoder which conditions on mel-spectrogram features, generated by a separate model. The iterative refinement process starts from Gaussian noise, and through a series of refinement steps (e.g., 50 steps), progressively recovers the audio sequence. WaveGrad 2 offers a natural way to trade-off between inference speed and sample quality, through adjusting the number of refinement steps. Experiments show that the model can generate high fidelity audio, approaching the performance of a state-of-the-art neural TTS system. We also report various ablation studies over different model configurations. Audio samples are available at https://wavegrad.github.io/v2.

FedNILM: Applying Federated Learning to NILM Applications at the Edge

Non-intrusive load monitoring (NILM) helps disaggregate the household's main electricity consumption to energy usages of individual appliances, thus greatly cutting down the cost in fine-grained household load monitoring. To address the arisen privacy concern in NILM applications, federated learning (FL) could be leveraged for NILM model training and sharing. When applying the FL paradigm in real-world NILM applications, however, we are faced with the challenges of edge resource restriction, edge model personalization and edge training data scarcity. In this paper we present FedNILM, a practical FL paradigm for NILM applications at the edge client. Specifically, FedNILM is designed to deliver privacy-preserving and personalized NILM services to large-scale edge clients, by leveraging i) secure data aggregation through federated learning, ii) efficient cloud model compression via filter pruning and multi-task learning, and iii) personalized edge model building with unsupervised transfer learning. Our experiments on real-world energy data show that, FedNILM is able to achieve personalized energy disaggregation with the state-of-the-art accuracy, while ensuring privacy preserving at the edge client.

Rethinking Training from Scratch for Object Detection

The ImageNet pre-training initialization is the de-facto standard for object detection. He et al. found it is possible to train detector from scratch(random initialization) while needing a longer training schedule with proper normalization technique. In this paper, we explore to directly pre-training on target dataset for object detection. Under this situation, we discover that the widely adopted large resizing strategy e.g. resize image to (1333, 800) is important for fine-tuning but it's not necessary for pre-training. Specifically, we propose a new training pipeline for object detection that follows `pre-training and fine-tuning', utilizing low resolution images within target dataset to pre-training detector then load it to fine-tuning with high resolution images. With this strategy, we can use batch normalization(BN) with large bath size during pre-training, it's also memory efficient that we can apply it on machine with very limited GPU memory(11G). We call it direct detection pre-training, and also use direct pre-training for short. Experiment results show that direct pre-training accelerates the pre-training phase by more than 11x on COCO dataset while with even +1.8mAP compared to ImageNet pre-training. Besides, we found direct pre-training is also applicable to transformer based backbones e.g. Swin Transformer. Code will be available.

More Behind Your Electricity Bill: a Dual-DNN Approach to Non-Intrusive Load Monitoring

Non-intrusive load monitoring (NILM) is a well-known single-channel blind source separation problem that aims to decompose the household energy consumption into itemised energy usage of individual appliances. In this way, considerable energy savings could be achieved by enhancing household's awareness of energy usage. Recent investigations have shown that deep neural networks (DNNs) based approaches are promising for the NILM task. Nevertheless, they normally ignore the inherent properties of appliance operations in the network design, potentially leading to implausible results. We are thus motivated to develop the dual Deep Neural Networks (dual-DNN), which aims to i) take advantage of DNNs' learning capability of latent features and ii) empower the DNN architecture with identification ability of universal properties. Specifically in the design of dual-DNN, we adopt one subnetwork to measure power ratings of different appliances' operation states, and the other subnetwork to identify the running states of target appliances. The final result is then obtained by multiplying these two network outputs and meanwhile considering the multi-state property of household appliances. To enforce the sparsity property in appliance's state operating, we employ median filtering and hard gating mechanisms to the subnetwork for state identification. Compared with the state-of-the-art NILM methods, our dual-DNN approach demonstrates a 21.67% performance improvement in average on two public benchmark datasets.

Scaling End-to-End Models for Large-Scale Multilingual ASR

Building ASR models across many language families is a challenging multi-task learning problem due to large language variations and heavily unbalanced data. Existing work has shown positive transfer from high resource to low resource languages. However, degradations on high resource languages are commonly observed due to interference from the heterogeneous multilingual data and reduction in per-language capacity. We conduct a capacity study on a 15-language task, with the amount of data per language varying from 7.7K to 54.7K hours. We adopt GShard [1] to efficiently scale up to 10B parameters. Empirically, we find that (1) scaling the number of model parameters is an effective way to solve the capacity bottleneck - our 500M-param model is already better than monolingual baselines and scaling it to 1B and 10B brought further quality gains; (2) larger models are not only more data efficient, but also more efficient in terms of training cost as measured in TPU days - the 1B-param model reaches the same accuracy at 34% of training time as the 500M-param model; (3) given a fixed capacity budget, adding depth usually works better than width and large encoders tend to do better than large decoders.

SpeechStew: Simply Mix All Available Speech Recognition Data to Train One Large Neural Network

We present SpeechStew, a speech recognition model that is trained on a combination of various publicly available speech recognition datasets: AMI, Broadcast News, Common Voice, LibriSpeech, Switchboard/Fisher, Tedlium, and Wall Street Journal. SpeechStew simply mixes all of these datasets together, without any special re-weighting or re-balancing of the datasets. SpeechStew achieves SoTA or near SoTA results across a variety of tasks, without the use of an external language model. Our results include 9.0\% WER on AMI-IHM, 4.7\% WER on Switchboard, 8.3\% WER on CallHome, and 1.3\% on WSJ, which significantly outperforms prior work with strong external language models. We also demonstrate that SpeechStew learns powerful transfer learning representations. We fine-tune SpeechStew on a noisy low resource speech dataset, CHiME-6. We achieve 38.9\% WER without a language model, which compares to 38.6\% WER to a strong HMM baseline with a language model.

Multi-Task Learning for End-to-End ASR Word and Utterance Confidence with Deletion Prediction

Confidence scores are very useful for downstream applications of automatic speech recognition (ASR) systems. Recent works have proposed using neural networks to learn word or utterance confidence scores for end-to-end ASR. In those studies, word confidence by itself does not model deletions, and utterance confidence does not take advantage of word-level training signals. This paper proposes to jointly learn word confidence, word deletion, and utterance confidence. Empirical results show that multi-task learning with all three objectives improves confidence metrics (NCE, AUC, RMSE) without the need for increasing the model size of the confidence estimation module. Using the utterance-level confidence for rescoring also decreases the word error rates on Google's Voice Search and Long-tail Maps datasets by 3-5% relative, without needing a dedicated neural rescorer.

Deep Latent Emotion Network for Multi-Task Learning

Feed recommendation models are widely adopted by numerous feed platforms to encourage users to explore the contents they are interested in. However, most of the current research simply focus on targeting user's preference and lack in-depth study of avoiding objectionable contents to be frequently recommended, which is a common reason that let user detest. To address this issue, we propose a Deep Latent Emotion Network (DLEN) model to extract latent probability of a user preferring a feed by modeling multiple targets with semi-supervised learning. With this method, the conflicts of different targets are successfully reduced in the training phase, which improves the training accuracy of each target effectively. Besides, by adding this latent state of user emotion to multi-target fusion, the model is capable of decreasing the probability to recommend objectionable contents to improve user retention and stay time during online testing phase. DLEN is deployed on a real-world multi-task feed recommendation scenario of Tencent QQ-Small-World with a dataset containing over a billion samples, and it exhibits a significant performance advantage over the SOTA MTL model in offline evaluation, together with a considerable increase by 3.02% in view-count and 2.63% in user stay-time in production. Complementary offline experiments of DLEN model on a public dataset also repeat improvements in various scenarios. At present, DLEN model has been successfully deployed in Tencent's feed recommendation system.

CSAFL: A Clustered Semi-Asynchronous Federated Learning Framework

Federated learning (FL) is an emerging distributed machine learning paradigm that protects privacy and tackles the problem of isolated data islands. At present, there are two main communication strategies of FL: synchronous FL and asynchronous FL. The advantages of synchronous FL are that the model has high precision and fast convergence speed. However, this synchronous communication strategy has the risk that the central server waits too long for the devices, namely, the straggler effect which has a negative impact on some time-critical applications. Asynchronous FL has a natural advantage in mitigating the straggler effect, but there are threats of model quality degradation and server crash. Therefore, we combine the advantages of these two strategies to propose a clustered semi-asynchronous federated learning (CSAFL) framework. We evaluate CSAFL based on four imbalanced federated datasets in a non-IID setting and compare CSAFL to the baseline methods. The experimental results show that CSAFL significantly improves test accuracy by more than +5% on the four datasets compared to TA-FedAvg. In particular, CSAFL improves absolute test accuracy by +34.4% on non-IID FEMNIST compared to TA-FedAvg.