The annotation for large-scale point clouds is still time-consuming and unavailable for many real-world tasks. Point cloud pre-training is one potential solution for obtaining a scalable model for fast adaptation. Therefore, in this paper, we investigate a new self-supervised learning approach, called Mixing and Disentangling (MD), for point cloud pre-training. As the name implies, we explore how to separate the original point cloud from the mixed point cloud, and leverage this challenging task as a pretext optimization objective for model training. Considering the limited training data in the original dataset, which is much less than prevailing ImageNet, the mixing process can efficiently generate more high-quality samples. We build one baseline network to verify our intuition, which simply contains two modules, encoder and decoder. Given a mixed point cloud, the encoder is first pre-trained to extract the semantic embedding. Then an instance-adaptive decoder is harnessed to disentangle the point clouds according to the embedding. Albeit simple, the encoder is inherently able to capture the point cloud keypoints after training and can be fast adapted to downstream tasks including classification and segmentation by the pre-training and fine-tuning paradigm. Extensive experiments on two datasets show that the encoder + ours (MD) significantly surpasses that of the encoder trained from scratch and converges quickly. In ablation studies, we further study the effect of each component and discuss the advantages of the proposed self-supervised learning strategy. We hope this self-supervised learning attempt on point clouds can pave the way for reducing the deeply-learned model dependence on large-scale labeled data and saving a lot of annotation costs in the future.
相關內容
Recently contrastive learning has shown significant progress in learning visual representations from unlabeled data. The core idea is training the backbone to be invariant to different augmentations of an instance. While most methods only maximize the feature similarity between two augmented data, we further generate more challenging training samples and force the model to keep predicting discriminative representation on these hard samples. In this paper, we propose MixSiam, a mixture-based approach upon the traditional siamese network. On the one hand, we input two augmented images of an instance to the backbone and obtain the discriminative representation by performing an element-wise maximum of two features. On the other hand, we take the mixture of these augmented images as input, and expect the model prediction to be close to the discriminative representation. In this way, the model could access more variant data samples of an instance and keep predicting invariant discriminative representations for them. Thus the learned model is more robust compared to previous contrastive learning methods. Extensive experiments on large-scale datasets show that MixSiam steadily improves the baseline and achieves competitive results with state-of-the-art methods. Our code will be released soon.
To date, most existing self-supervised learning methods are designed and optimized for image classification. These pre-trained models can be sub-optimal for dense prediction tasks due to the discrepancy between image-level prediction and pixel-level prediction. To fill this gap, we aim to design an effective, dense self-supervised learning method that directly works at the level of pixels (or local features) by taking into account the correspondence between local features. We present dense contrastive learning, which implements self-supervised learning by optimizing a pairwise contrastive (dis)similarity loss at the pixel level between two views of input images. Compared to the baseline method MoCo-v2, our method introduces negligible computation overhead (only <1% slower), but demonstrates consistently superior performance when transferring to downstream dense prediction tasks including object detection, semantic segmentation and instance segmentation; and outperforms the state-of-the-art methods by a large margin. Specifically, over the strong MoCo-v2 baseline, our method achieves significant improvements of 2.0% AP on PASCAL VOC object detection, 1.1% AP on COCO object detection, 0.9% AP on COCO instance segmentation, 3.0% mIoU on PASCAL VOC semantic segmentation and 1.8% mIoU on Cityscapes semantic segmentation. Code is available at: //git.io/AdelaiDet
The pre-training models such as BERT have achieved great results in various natural language processing problems. However, a large number of parameters need significant amounts of memory and the consumption of inference time, which makes it difficult to deploy them on edge devices. In this work, we propose a knowledge distillation method LRC-BERT based on contrastive learning to fit the output of the intermediate layer from the angular distance aspect, which is not considered by the existing distillation methods. Furthermore, we introduce a gradient perturbation-based training architecture in the training phase to increase the robustness of LRC-BERT, which is the first attempt in knowledge distillation. Additionally, in order to better capture the distribution characteristics of the intermediate layer, we design a two-stage training method for the total distillation loss. Finally, by verifying 8 datasets on the General Language Understanding Evaluation (GLUE) benchmark, the performance of the proposed LRC-BERT exceeds the existing state-of-the-art methods, which proves the effectiveness of our method.
One paradigm for learning from few labeled examples while making best use of a large amount of unlabeled data is unsupervised pretraining followed by supervised fine-tuning. Although this paradigm uses unlabeled data in a task-agnostic way, in contrast to common approaches to semi-supervised learning for computer vision, we show that it is surprisingly effective for semi-supervised learning on ImageNet. A key ingredient of our approach is the use of big (deep and wide) networks during pretraining and fine-tuning. We find that, the fewer the labels, the more this approach (task-agnostic use of unlabeled data) benefits from a bigger network. After fine-tuning, the big network can be further improved and distilled into a much smaller one with little loss in classification accuracy by using the unlabeled examples for a second time, but in a task-specific way. The proposed semi-supervised learning algorithm can be summarized in three steps: unsupervised pretraining of a big ResNet model using SimCLRv2, supervised fine-tuning on a few labeled examples, and distillation with unlabeled examples for refining and transferring the task-specific knowledge. This procedure achieves 73.9% ImageNet top-1 accuracy with just 1% of the labels ($\le$13 labeled images per class) using ResNet-50, a $10\times$ improvement in label efficiency over the previous state-of-the-art. With 10% of labels, ResNet-50 trained with our method achieves 77.5% top-1 accuracy, outperforming standard supervised training with all of the labels.
Contrastive learning has become a key component of self-supervised learning approaches for computer vision. By learning to embed two augmented versions of the same image close to each other and to push the embeddings of different images apart, one can train highly transferable visual representations. As revealed by recent studies, heavy data augmentation and large sets of negatives are both crucial in learning such representations. At the same time, data mixing strategies either at the image or the feature level improve both supervised and semi-supervised learning by synthesizing novel examples, forcing networks to learn more robust features. In this paper, we argue that an important aspect of contrastive learning, i.e., the effect of hard negatives, has so far been neglected. To get more meaningful negative samples, current top contrastive self-supervised learning approaches either substantially increase the batch sizes, or keep very large memory banks; increasing the memory size, however, leads to diminishing returns in terms of performance. We therefore start by delving deeper into a top-performing framework and show evidence that harder negatives are needed to facilitate better and faster learning. Based on these observations, and motivated by the success of data mixing, we propose hard negative mixing strategies at the feature level, that can be computed on-the-fly with a minimal computational overhead. We exhaustively ablate our approach on linear classification, object detection and instance segmentation and show that employing our hard negative mixing procedure improves the quality of visual representations learned by a state-of-the-art self-supervised learning method.
Increasing model size when pretraining natural language representations often results in improved performance on downstream tasks. However, at some point further model increases become harder due to GPU/TPU memory limitations, longer training times, and unexpected model degradation. To address these problems, we present two parameter-reduction techniques to lower memory consumption and increase the training speed of BERT. Comprehensive empirical evidence shows that our proposed methods lead to models that scale much better compared to the original BERT. We also use a self-supervised loss that focuses on modeling inter-sentence coherence, and show it consistently helps downstream tasks with multi-sentence inputs. As a result, our best model establishes new state-of-the-art results on the GLUE, RACE, and SQuAD benchmarks while having fewer parameters compared to BERT-large.
Language model pre-training, such as BERT, has significantly improved the performances of many natural language processing tasks. However, pre-trained language models are usually computationally expensive and memory intensive, so it is difficult to effectively execute them on some resource-restricted devices. To accelerate inference and reduce model size while maintaining accuracy, we firstly propose a novel transformer distillation method that is a specially designed knowledge distillation (KD) method for transformer-based models. By leveraging this new KD method, the plenty of knowledge encoded in a large teacher BERT can be well transferred to a small student TinyBERT. Moreover, we introduce a new two-stage learning framework for TinyBERT, which performs transformer distillation at both the pre-training and task-specific learning stages. This framework ensures that TinyBERT can capture both the general-domain and task-specific knowledge of the teacher BERT. TinyBERT is empirically effective and achieves comparable results with BERT in GLUE datasets, while being 7.5x smaller and 9.4x faster on inference. TinyBERT is also significantly better than state-of-the-art baselines, even with only about 28% parameters and 31% inference time of baselines.
We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models, BERT is designed to pre-train deep bidirectional representations by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT representations can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE benchmark to 80.4% (7.6% absolute improvement), MultiNLI accuracy to 86.7 (5.6% absolute improvement) and the SQuAD v1.1 question answering Test F1 to 93.2 (1.5% absolute improvement), outperforming human performance by 2.0%.
Video representation learning is a vital problem for classification task. Recently, a promising unsupervised paradigm termed self-supervised learning has emerged, which explores inherent supervisory signals implied in massive data for feature learning via solving auxiliary tasks. However, existing methods in this regard suffer from two limitations when extended to video classification. First, they focus only on a single task, whereas ignoring complementarity among different task-specific features and thus resulting in suboptimal video representation. Second, high computational and memory cost hinders their application in real-world scenarios. In this paper, we propose a graph-based distillation framework to address these problems: (1) We propose logits graph and representation graph to transfer knowledge from multiple self-supervised tasks, where the former distills classifier-level knowledge by solving a multi-distribution joint matching problem, and the latter distills internal feature knowledge from pairwise ensembled representations with tackling the challenge of heterogeneity among different features; (2) The proposal that adopts a teacher-student framework can reduce the redundancy of knowledge learnt from teachers dramatically, leading to a lighter student model that solves classification task more efficiently. Experimental results on 3 video datasets validate that our proposal not only helps learn better video representation but also compress model for faster inference.
Existing manifold learning methods are not appropriate for image retrieval task, because most of them are unable to process query image and they have much additional computational cost especially for large scale database. Therefore, we propose the iterative manifold embedding (IME) layer, of which the weights are learned off-line by unsupervised strategy, to explore the intrinsic manifolds by incomplete data. On the large scale database that contains 27000 images, IME layer is more than 120 times faster than other manifold learning methods to embed the original representations at query time. We embed the original descriptors of database images which lie on manifold in a high dimensional space into manifold-based representations iteratively to generate the IME representations in off-line learning stage. According to the original descriptors and the IME representations of database images, we estimate the weights of IME layer by ridge regression. In on-line retrieval stage, we employ the IME layer to map the original representation of query image with ignorable time cost (2 milliseconds). We experiment on five public standard datasets for image retrieval. The proposed IME layer significantly outperforms related dimension reduction methods and manifold learning methods. Without post-processing, Our IME layer achieves a boost in performance of state-of-the-art image retrieval methods with post-processing on most datasets, and needs less computational cost.
小貼士
登錄享
相關主題
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191