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.
The recent development of the on-chip micro-polarizer technology has made it possible to acquire four spatially aligned and temporally synchronized polarization images with the same ease of operation as a conventional camera. In this paper, we investigate the use of this sensor technology in high-dynamic-range (HDR) imaging. Specifically, observing that natural light can be attenuated differently by varying the orientation of the polarization filter, we treat the multiple images captured by the polarization camera as a set captured under different exposure times. In our approach, we first study the relationship among polarizer orientation, degree and angle of polarization of light to the exposure time of a pixel in the polarization image. Subsequently, we propose a deep snapshot HDR reconstruction framework to recover an HDR image using the polarization images. A polarized HDR dataset is created to train and evaluate our approach. We demonstrate that our approach performs favorably against state-of-the-art HDR reconstruction algorithms.
In practice, the problems encountered in training NAS (Neural Architecture Search) are not simplex, but a series of combinations of difficulties are often faced(incorrect compensation estimation, curse of dimension, overfitting, high complexity, etc.). From the point of view for solving practical problems, this paper makes reference and improvement to the previous researches which only solve the single problem of NAS, and combines them into a practical technology flow. This paper propose a framework that decouples the network structure from the search space for operators. We use two BOHBs(Bayesian Optimization Hyperband) to search alternately in the vast network structure and operator search space. And then, we trained a GCN-baesd predictor using the feedback of the child model. This approach takes care of the dimension curse while improving efficiency. Considering that activation function and initialization are also important components of neural network, and can affect the generalization ability of the model. This paper introduced an activation function and an initialization method domain, join them to the operator search space to form a generalized search space, thus improving the generalization ability of the child model. At last, We applied our framework to neural architecture search and achieved significant improvements on multiple datasets.
With the development of computer-aided diagnosis (CAD) and image scanning technology, Whole-slide Image (WSI) scanners are widely used in the field of pathological diagnosis. Therefore, WSI analysis has become the key to modern digital pathology. Since 2004, WSI has been used more and more in CAD. Since machine vision methods are usually based on semi-automatic or fully automatic computers, they are highly efficient and labor-saving. The combination of WSI and CAD technologies for segmentation, classification, and detection helps histopathologists obtain more stable and quantitative analysis results, save labor costs and improve diagnosis objectivity. This paper reviews the methods of WSI analysis based on machine learning. Firstly, the development status of WSI and CAD methods are introduced. Secondly, we discuss publicly available WSI datasets and evaluation metrics for segmentation, classification, and detection tasks. Then, the latest development of machine learning in WSI segmentation, classification, and detection are reviewed continuously. Finally, the existing methods are studied, the applicabilities of the analysis methods are analyzed, and the application prospects of the analysis methods in this field are forecasted.
This paper is concerned with polarimetric dense map reconstruction based on a polarization camera with the help of relative depth information as a prior. In general, polarization imaging is able to reveal information about surface normal such as azimuth and zenith angles, which can support the development of solutions to the problem of dense reconstruction, especially in texture-poor regions. However, polarimetric shape cues are ambiguous due to two types of polarized reflection (specular/diffuse). Although methods have been proposed to address this issue, they either are offline and therefore not practical in robotics applications, or use incomplete polarimetric cues, leading to sub-optimal performance. In this paper, we propose an online reconstruction method that uses full polarimetric cues available from the polarization camera. With our online method, we can propagate sparse depth values both along and perpendicular to iso-depth contours. Through comprehensive experiments on challenging image sequences, we demonstrate that our method is able to significantly improve the accuracy of the depthmap as well as increase its density, specially in regions of poor texture.
A Multistage Full Matching disparity estimation scheme (MFM) is proposed in this work. We demonstrate that decouple all similarity scores directly from the low-resolution 4D volume step by step instead of estimating low-resolution 3D cost volume through focusing on optimizing the low-resolution 4D volume iteratively leads to more accurate disparity. To this end, we first propose to decompose the full matching task into multiple stages of the cost aggregation module. Specifically, we decompose the high-resolution predicted results into multiple groups, and every stage of the newly designed cost aggregation module learns only to estimate the results for a group of points. This alleviates the problem of feature internal competitive when learning similarity scores of all candidates from one low-resolution 4D volume output from one stage. Then, we propose the strategy of \emph{Stages Mutual Aid}, which takes advantage of the relationship of multiple stages to boost similarity scores estimation of each stage, to solve the unbalanced prediction of multiple stages caused by serial multistage framework. Experiment results demonstrate that the proposed method achieves more accurate disparity estimation results and outperforms state-of-the-art methods on Scene Flow, KITTI 2012 and KITTI 2015 datasets.
Being a crucial task of autonomous driving, Stereo matching has made great progress in recent years. Existing stereo matching methods estimate disparity instead of depth. They treat the disparity errors as the evaluation metric of the depth estimation errors, since the depth can be calculated from the disparity according to the triangulation principle. However, we find that the error of the depth depends not only on the error of the disparity but also on the depth range of the points. Therefore, even if the disparity error is low, the depth error is still large, especially for the distant points. In this paper, a novel Direct Depth Learning Network (DDL-Net) is designed for stereo matching. DDL-Net consists of two stages: the Coarse Depth Estimation stage and the Adaptive-Grained Depth Refinement stage, which are all supervised by depth instead of disparity. Specifically, Coarse Depth Estimation stage uniformly samples the matching candidates according to depth range to construct cost volume and output coarse depth. Adaptive-Grained Depth Refinement stage performs further matching near the coarse depth to correct the imprecise matching and wrong matching. To make the Adaptive-Grained Depth Refinement stage robust to the coarse depth and adaptive to the depth range of the points, the Granularity Uncertainty is introduced to Adaptive-Grained Depth Refinement stage. Granularity Uncertainty adjusts the matching range and selects the candidates' features according to coarse prediction confidence and depth range. We verify the performance of DDL-Net on SceneFlow dataset and DrivingStereo dataset by different depth metrics. Results show that DDL-Net achieves an average improvement of 25% on the SceneFlow dataset and $12\%$ on the DrivingStereo dataset comparing the classical methods. More importantly, we achieve state-of-the-art accuracy at a large distance.
Image-to-image translation aims at translating a particular style of an image to another. The synthesized images can be more photo-realistic and identity-preserving by decomposing the image into content and style in a disentangled manner. While existing models focus on designing specialized network architecture to separate the two components, this paper investigates how to explicitly constrain the content and style statistics of images. We achieve this goal by transforming the input image into high frequency and low frequency information, which correspond to the content and style, respectively. We regulate the frequency distribution from two aspects: a) a spatial level restriction to locally restrict the frequency distribution of images; b) a spectral level regulation to enhance the global consistency among images. On multiple datasets we show that the proposed approach consistently leads to significant improvements on top of various state-of-the-art image translation models.
We present an Object-aware Feature Aggregation (OFA) module for video object detection (VID). Our approach is motivated by the intriguing property that video-level object-aware knowledge can be employed as a powerful semantic prior to help object recognition. As a consequence, augmenting features with such prior knowledge can effectively improve the classification and localization performance. To make features get access to more content about the whole video, we first capture the object-aware knowledge of proposals and incorporate such knowledge with the well-established pair-wise contexts. With extensive experimental results on the ImageNet VID dataset, our approach demonstrates the effectiveness of object-aware knowledge with the superior performance of 83.93% and 86.09% mAP with ResNet-101 and ResNeXt-101, respectively. When further equipped with Sequence DIoU NMS, we obtain the best-reported mAP of 85.07% and 86.88% upon the paper submitted. The code to reproduce our results will be released after acceptance.