As a promising scheme of self-supervised learning, masked autoencoding has significantly advanced natural language processing and computer vision. Inspired by this, we propose a neat scheme of masked autoencoders for point cloud self-supervised learning, addressing the challenges posed by point cloud's properties, including leakage of location information and uneven information density. Concretely, we divide the input point cloud into irregular point patches and randomly mask them at a high ratio. Then, a standard Transformer based autoencoder, with an asymmetric design and a shifting mask tokens operation, learns high-level latent features from unmasked point patches, aiming to reconstruct the masked point patches. Extensive experiments show that our approach is efficient during pre-training and generalizes well on various downstream tasks. Specifically, our pre-trained models achieve 85.18% accuracy on ScanObjectNN and 94.04% accuracy on ModelNet40, outperforming all the other self-supervised learning methods. We show with our scheme, a simple architecture entirely based on standard Transformers can surpass dedicated Transformer models from supervised learning. Our approach also advances state-of-the-art accuracies by 1.5%-2.3% in the few-shot object classification. Furthermore, our work inspires the feasibility of applying unified architectures from languages and images to the point cloud.
Sensor-based environmental perception is a crucial part of the autonomous driving system. In order to get an excellent perception of the surrounding environment, an intelligent system would configure multiple LiDARs (3D Light Detection and Ranging) to cover the distant and near space of the car. The precision of perception relies on the quality of sensor calibration. This research aims at developing an accurate, automatic, and robust calibration strategy for multiple LiDAR systems in the general road scene. We thus propose CROON (automatiC multi-LiDAR CalibratiOn and Refinement method in rOad sceNe), a two-stage method including rough and refinement calibration. The first stage can calibrate the sensor from an arbitrary initial pose, and the second stage is able to precisely calibrate the sensor iteratively. Specifically, CROON utilize the nature characteristics of road scene so that it is independent and easy to apply in large-scale conditions. Experimental results on real-world and simulated data sets demonstrate the reliability and accuracy of our method. All the related data sets and codes are open-sourced on the Github website https://github.com/OpenCalib/LiDAR2LiDAR.
Training a text-to-image generator in the general domain (e.g., Dall.e, CogView) requires huge amounts of paired text-image data, which is too expensive to collect. In this paper, we propose a self-supervised scheme named as CLIP-GEN for general text-to-image generation with the language-image priors extracted with a pre-trained CLIP model. In our approach, we only require a set of unlabeled images in the general domain to train a text-to-image generator. Specifically, given an image without text labels, we first extract the embedding of the image in the united language-vision embedding space with the image encoder of CLIP. Next, we convert the image into a sequence of discrete tokens in the VQGAN codebook space (the VQGAN model can be trained with the unlabeled image dataset in hand). Finally, we train an autoregressive transformer that maps the image tokens from its unified language-vision representation. Once trained, the transformer can generate coherent image tokens based on the text embedding extracted from the text encoder of CLIP upon an input text. Such a strategy enables us to train a strong and general text-to-image generator with large text-free image dataset such as ImageNet. Qualitative and quantitative evaluations verify that our method significantly outperforms optimization-based text-to-image methods in terms of image quality while not compromising the text-image matching. Our method can even achieve comparable performance as flagship supervised models like CogView.
Neural image compression have reached or out-performed traditional methods (such as JPEG, BPG, WebP). However,their sophisticated network structures with cascaded convolution layers bring heavy computational burden for practical deployment. In this paper, we explore the structural sparsity in neural image compression network to obtain real-time acceleration without any specialized hardware design or algorithm. We propose a simple plug-in adaptive binary channel masking(ABCM) to judge the importance of each convolution channel and introduce sparsity during training. During inference, the unimportant channels are pruned to obtain slimmer network and less computation. We implement our method into three neural image compression networks with different entropy models to verify its effectiveness and generalization, the experiment results show that up to 7x computation reduction and 3x acceleration can be achieved with negligible performance drop.
Single image deraining has been an important topic in low-level computer vision tasks. The atmospheric veiling effect (which is generated by rain accumulation, similar to fog) usually appears with the rain. Most deep learning-based single image deraining methods mainly focus on rain streak removal by disregarding this effect, which leads to low-quality deraining performance. In addition, these methods are trained only on synthetic data, hence they do not take into account real-world rainy images. To address the above issues, we propose a novel unsupervised attentive-adversarial learning framework (UALF) for single image deraining that trains on both synthetic and real rainy images while simultaneously capturing both rain streaks and rain accumulation features. UALF consists of a Rain-fog2Clean (R2C) transformation block and a Clean2Rain-fog (C2R) transformation block. In R2C, to better characterize the rain-fog fusion feature and to achieve high-quality deraining performance, we employ an attention rain-fog feature extraction network (ARFE) to exploit the self-similarity of global and local rain-fog information by learning the spatial feature correlations. Moreover, to improve the transformation ability of C2R, we design a rain-fog feature decoupling and reorganization network (RFDR) by embedding a rainy image degradation model and a mixed discriminator to preserve richer texture details. Extensive experiments on benchmark rain-fog and rain datasets show that UALF outperforms state-of-the-art deraining methods. We also conduct defogging performance evaluation experiments to further demonstrate the effectiveness of UALF
Adversarial learning-based image defogging methods have been extensively studied in computer vision due to their remarkable performance. However, most existing methods have limited defogging capabilities for real cases because they are trained on the paired clear and synthesized foggy images of the same scenes. In addition, they have limitations in preserving vivid color and rich textual details in defogging. To address these issues, we develop a novel generative adversarial network, called holistic attention-fusion adversarial network (HAAN), for single image defogging. HAAN consists of a Fog2Fogfree block and a Fogfree2Fog block. In each block, there are three learning-based modules, namely, fog removal, color-texture recovery, and fog synthetic, that are constrained each other to generate high quality images. HAAN is designed to exploit the self-similarity of texture and structure information by learning the holistic channel-spatial feature correlations between the foggy image with its several derived images. Moreover, in the fog synthetic module, we utilize the atmospheric scattering model to guide it to improve the generative quality by focusing on an atmospheric light optimization with a novel sky segmentation network. Extensive experiments on both synthetic and real-world datasets show that HAAN outperforms state-of-the-art defogging methods in terms of quantitative accuracy and subjective visual quality.
Recently, MLP-like vision models have achieved promising performances on mainstream visual recognition tasks. In contrast with vision transformers and CNNs, the success of MLP-like models shows that simple information fusion operations among tokens and channels can yield a good representation power for deep recognition models. However, existing MLP-like models fuse tokens through static fusion operations, lacking adaptability to the contents of the tokens to be mixed. Thus, customary information fusion procedures are not effective enough. To this end, this paper presents an efficient MLP-like network architecture, dubbed DynaMixer, resorting to dynamic information fusion. Critically, we propose a procedure, on which the DynaMixer model relies, to dynamically generate mixing matrices by leveraging the contents of all the tokens to be mixed. To reduce the time complexity and improve the robustness, a dimensionality reduction technique and a multi-segment fusion mechanism are adopted. Our proposed DynaMixer model (97M parameters) achieves 84.3\% top-1 accuracy on the ImageNet-1K dataset without extra training data, performing favorably against the state-of-the-art vision MLP models. When the number of parameters is reduced to 26M, it still achieves 82.7\% top-1 accuracy, surpassing the existing MLP-like models with a similar capacity. The implementation of DynaMixer will be made available to the public.