기법

DBlock is a residual based block used in the discriminator of the GAN-TTS architecture. They are similar to the GBlocks used in the generator, but without batch normalisation.

DELG is a convolutional neural network for image retrieval that combines generalized mean pooling for global features and attentive selection for local features. The entire network can be learned end-to-end by carefully balancing the gradient flow between two heads – requiring only image-level labels. This allows for efficient inference by extracting an image’s global feature, detected keypoints and local descriptors within a single model. The model is enabled by leveraging hierarchical image representations that arise in CNNs, which are coupled to generalized mean pooling and attentive local feature detection. Secondly, a convolutional autoencoder module is adopted that can successfully learn low-dimensional local descriptors. This can be readily integrated into the unified model, and avoids the need of post-processing learning steps, such as PCA, that are commonly used. Finally, a procedure is used that enables end-to-end training of the proposed model using only image-level supervision. This requires carefully controlling the gradient flow between the global and local network heads during backpropagation, to avoid disrupting the desired representations.

The DELU is a type of activation function that has trainable parameters, uses the complex linear and exponential functions in the positive dimension and uses the SiLU in the negative dimension. $$DELU(x) = SiLU(x), x \leqslant 0$$ $$DELU(x) = (n + 0.5)x + |e^{-x} - 1|, x > 0$$

DIoU-NMS is a type of non-maximum suppression where we use Distance IoU rather than regular DIoU, in which the overlap area and the distance between two central points of bounding boxes are simultaneously considered when suppressing redundant boxes. In original NMS, the IoU metric is used to suppress the redundant detection boxes, where the overlap area is the unique factor, often yielding false suppression for the cases with occlusion. With DIoU-NMS, we not only consider the overlap area but also central point distance between two boxes.

DNN2LR is an automatic feature crossing method to find feature interactions in a deep neural network, and use them as cross features in logistic regression. In general, DNN2LR consists of two steps: (1) generating a compact and accurate candidate set of cross feature fields; (2) searching in the candidate set for the final cross feature fields.

In the field of fusing multi-spectral and panchromatic images (Pan-sharpening), the impressive effectiveness of deep neural networks has been recently employed to overcome the drawbacks of traditional linear models and boost the fusing accuracy. However, to the best of our knowledge, existing research works are mainly based on simple and flat networks with relatively shallow architecture, which severely limited their performances. In this paper, the concept of residual learning has been introduced to form a very deep convolutional neural network to make a full use of the high non-linearity of deep learning models. By both quantitative and visual assessments on a large number of high quality multi-spectral images from various sources, it has been supported that our proposed model is superior to all mainstream algorithms included in the comparison, and achieved the highest spatial-spectral unified accuracy.

DSelect-k is a continuously differentiable and sparse gate for Mixture-of-experts (MoE), based on a novel binary encoding formulation. Given a user-specified parameter $k$, the gate selects at most $k$ out of the $n$ experts. The gate can be trained using first-order methods, such as stochastic gradient descent, and offers explicit control over the number of experts to select. This explicit control over sparsity leads to a cardinality-constrained optimization problem, which is computationally challenging. To circumvent this challenge, the authors use a unconstrained reformulation that is equivalent to the original problem. The reformulated problem uses a binary encoding scheme to implicitly enforce the cardinality constraint. By carefully smoothing the binary encoding variables, the reformulated problem can be effectively optimized using first-order methods such as SGD. The motivation for this method is that existing sparse gates, such as Top-k, are not smooth. The lack of smoothness can lead to convergence and statistical performance issues when training with gradient-based methods.

DVD-GAN DBlock is a residual block for the discriminator used in the DVD-GAN architecture for video generation. Unlike regular residual blocks, 3D convolutions are employed due to the application to multiple frames in a video.

DecomCAM decomposes intermediate activation maps into orthogonal features using singular value decomposition and generates saliency maps by integrating them.

DeepViT is a type of vision transformer that replaces the self-attention layer within the transformer block with a Re-attention module to address the issue of attention collapse and enables training deeper ViTs.

Deflation is a video-to-image operation to transform a video network into a network that can ingest a single image. In the two types of video networks considered in the original paper, this deflation corresponds to the following operations: for 3D convolutional based networks, summing the 3D spatio-temporal filters over the temporal dimension to obtain 2D filters; for TSM networks,, turning off the channel shifting which results in a standard residual architecture (ResNet50) for images.

Deformable ConvNets do not learn an affine transformation. They divide convolution into two steps, firstly sampling features on a regular grid $ \mathcal{R} $ from the input feature map, then aggregating sampled features by weighted summation using a convolution kernel. The process can be written as: \begin{align} Y(p_{0}) &= \sum_{p_i \in \mathcal{R}} w(p_{i}) X(p_{0} + p_{i}) \end{align} \begin{align} \mathcal{R} &= {(-1,-1), (-1, 0), \dots, (1, 1)} \end{align} The deformable convolution augments the sampling process by introducing a group of learnable offsets $\Delta p_{i}$ which can be generated by a lightweight CNN. Using the offsets $\Delta p_{i}$, the deformable convolution can be formulated as: \begin{align} Y(p_{0}) &= \sum_{p_i \in \mathcal{R}} w(p_{i}) X(p_{0} + p_{i} + \Delta p_{i}). \end{align} Through the above method, adaptive sampling is achieved. However, $\Delta p_{i}$ is a floating point value unsuited to grid sampling. To address this problem, bilinear interpolation is used. Deformable RoI pooling is also used, which greatly improves object detection. Deformable ConvNets adaptively select the important regions and enlarge the valid receptive field of convolutional neural networks; this is important in object detection and semantic segmentation tasks.

DenseNet-Elastic is a convolutional neural network that is a modification of a DenseNet with elastic blocks (extra upsampling and downsampling).

Discriminative Adversarial Search, or DAS, is a sequence decoding approach which aims to alleviate the effects of exposure bias and to optimize on the data distribution itself rather than for external metrics. Inspired by generative adversarial networks (GANs), wherein a discriminator is used to improve the generator, DAS differs from GANs in that the generator parameters are not updated at training time and the discriminator is only used to drive sequence generation at inference time.

Drafting Network is a style transfer module designed to transfer global style patterns in low-resolution, since global patterns can be transferred easier in low resolution due to larger receptive field and less local details. To achieve single style transfer, earlier work trained an encoder-decoder module, where only the content image is used as input. To better combine the style feature and the content feature, the Drafting Network adopts the AdaIN module. The architecture of Drafting Network is shown in the Figure, which includes an encoder, several AdaIN modules and a decoder. (1) The encoder is a pre-trained VGG-19 network, which is fixed during training. Given $\bar{x}\_{c}$ and $\bar{x}\_{s}$, the VGG encoder extracts features in multiple granularity at 2_1, 3_1 and 4_1 layers. (2) Then, we apply feature modulation between the content and style feature using AdaIN modules after 2_1, 3_1 and 4_1 layers, respectively. (3) Finally, in each granularity of decoder, the corresponding feature from the AdaIN module is merged via a skip-connection. Here, skip-connections after AdaIN modules in both low and high levels are leveraged to help to reserve content structure, especially for low-resolution image.

Contrastive learning has achieved remarkable success in representation learning via self-supervision in unsupervised settings. However, effectively adapting contrastive learning to supervised learning tasks remains as a challenge in practice. In this work, we introduce a dual contrastive learning (DualCL) framework that simultaneously learns the features of input samples and the parameters of classifiers in the same space. Specifically, DualCL regards the parameters of the classifiers as augmented samples associating to different labels and then exploits the contrastive learning between the input samples and the augmented samples. Empirical studies on five benchmark text classification datasets and their low-resource version demonstrate the improvement in classification accuracy and confirm the capability of learning discriminative representations of DualCL.

DynaBERT is a BERT-variant which can flexibly adjust the size and latency by selecting adaptive width and depth. The training process of DynaBERT includes first training a width-adaptive BERT and then allowing both adaptive width and depth, by distilling knowledge from the full-sized model to small sub-networks. Network rewiring is also used to keep the more important attention heads and neurons shared by more sub-networks. A two-stage procedure is used to train DynaBERT. First, using knowledge distillation (dashed lines) to transfer the knowledge from a fixed teacher model to student sub-networks with adaptive width in DynaBERTW. Then, using knowledge distillation (dashed lines) to transfer the knowledge from a trained DynaBERTW to student sub-networks with adaptive width and depth in DynaBERT.

Entropy Minimized Ensemble of Adapters, or EMEA, is a method that optimizes the ensemble weights of the pretrained language adapters for each test sentence by minimizing the entropy of its predictions. The intuition behind the method is that a good adapter weight $\alpha$ for a test input $x$ should make the model more confident in its prediction for $x$, that is, it should lead to lower model entropy over the input

Introduced by Hinton et al. in 2012, dropout has stood the test of time as a regularizer for preventing overfitting in neural networks. In this study, we demonstrate that dropout can also mitigate underfitting when used at the start of training. During the early phase, we find dropout reduces the directional variance of gradients across mini-batches and helps align the mini-batch gradients with the entire dataset's gradient. This helps counteract the stochasticity of SGD and limit the influence of individual batches on model training. Our findings lead us to a solution for improving performance in underfitting models - early dropout: dropout is applied only during the initial phases of training, and turned off afterwards. Models equipped with early dropout achieve lower final training loss compared to their counterparts without dropout. Additionally, we explore a symmetric technique for regularizing overfitting models - late dropout, where dropout is not used in the early iterations and is only activated later in training. Experiments on ImageNet and various vision tasks demonstrate that our methods consistently improve generalization accuracy. Our results encourage more research on understanding regularization in deep learning and our methods can be useful tools for future neural network training, especially in the era of large data. Code is available at https://github.com/facebookresearch/dropout .

EdgeBoxes is an approach for generating object bounding box proposals directly from edges. Similar to segments, edges provide a simplified but informative representation of an image. In fact, line drawings of an image can accurately convey the high-level information contained in an image using only a small fraction of the information. The main insight behind the method is the observation: the number of contours wholly enclosed by a bounding box is indicative of the likelihood of the box containing an object. We say a contour is wholly enclosed by a box if all edge pixels belonging to the contour lie within the interior of the box. Edges tend to correspond to object boundaries, and as such boxes that tightly enclose a set of edges are likely to contain an object. However, some edges that lie within an object’s bounding box may not be part of the contained object. Specifically, edge pixels that belong to contours straddling the box’s boundaries are likely to correspond to objects or structures that lie outside the box. Source: Zitnick and Dollar

Elastic Dense Block is a skip connection block that modifies the Dense Block with downsamplings and upsamplings in parallel branches at each layer to let the network learn from a data scaling policy in which inputs are processed at different resolutions in each layer. It is called "elastic" because each layer in the network is flexible in terms of choosing the best scale by a soft policy.

An End-to-End Memory Network is a neural network with a recurrent attention model over a possibly large external memory. The architecture is a form of Memory Network, but unlike the model in that work, it is trained end-to-end, and hence requires significantly less supervision during training. It can also be seen as an extension of RNNsearch to the case where multiple computational steps (hops) are performed per output symbol. The model takes a discrete set of inputs $x\_{1}, \dots, x\_{n}$ that are to be stored in the memory, a query $q$, and outputs an answer $a$. Each of the $x\_{i}$, $q$, and $a$ contains symbols coming from a dictionary with $V$ words. The model writes all $x$ to the memory up to a fixed buffer size, and then finds a continuous representation for the $x$ and $q$. The continuous representation is then processed via multiple hops to output $a$.

Energy Based Processes extend energy based models to exchangeable data while allowing neural network parameterizations of the energy function. They extend the previously separate stochastic process and latent variable model perspectives in a common framework. The result is a generalization of Gaussian processes and Student-t processes that exploits EBMs for greater flexibility.

ExtremeNet is a a bottom-up object detection framework that detects four extreme points (top-most, left-most, bottom-most, right-most) of an object. It uses a keypoint estimation framework to find extreme points, by predicting four multi-peak heatmaps for each object category. In addition, it uses one heatmap per category predicting the object center, as the average of two bounding box edges in both the x and y dimension. We group extreme points into objects with a purely geometry-based approach. We group four extreme points, one from each map, if and only if their geometric center is predicted in the center heatmap with a score higher than a pre-defined threshold, We enumerate all $O\left(n^{4}\right)$ combinations of extreme point prediction, and select the valid ones.

Feature Fusion Module v1 is a feature fusion module from the M2Det object detection model, and feature fusion modules are crucial for constructing the final multi-level feature pyramid. They use 1x1 convolution layers to compress the channels of the input features and use concatenation operation to aggregate these feature map. FFMv1 takes two feature maps with different scales in backbone as input, it adopts one upsample operation to rescale the deep features to the same scale before the concatenation operation.

Feature Fusion Module v2 is a feature fusion module from the M2Det object detection model, and is crucial for constructing the final multi-level feature pyramid. They use 1x1 convolution layers to compress the channels of the input features and use a concatenation operation to aggregate these feature map. FFMv2 takes the base feature and the largest output feature map of the previous Thinned U-Shape Module (TUM) – these two are of the same scale – as input, and produces the fused feature for the next TUM.

FINCH is a parameter-free fast and scalable clustering algorithm. it stands out for its speed and clustering quality.

A variant of CutMix which randomly samples masks from Fourier space.

The motive for ensembling is to utilize each of the confidence factors generated from base learners fully by mapping them into non-linear functions. One of the mapped values signifies the abidance or closeness to 1 and the other one signifies the deviation from 1. This proposed approach overcomes the shortcoming of the conventional ranking methods. The scores from base learners are mapped on two different functions having different concavities to generate non-linear fuzzy ranks and generate a fused score by combining these two ranks, which helps us to quantify the total deviation from expected. Lesser the deviation shows better confidence towards a particular class. The class having the lowest deviation value is considered as the winner and is assigned as the final class value. Here, we first give a brief overview of the pre-trained CNN models used as base learners.

FashionCLIP is a fine-tuned CLIP model on fashion data (more than 800K pairs). It is the first foundation model for Fashion.

Fast-OCR is a new lightweight detection network that incorporates features from existing models focused on the speed/accuracy trade-off, such as YOLOv2, CR-NET, and Fast-YOLOv4.

The Fast-YOLOv4-SmallObj model is a modified version of Fast-YOLOv4 to improve the detection of small objects. Seven layers were added so that it predicts bounding boxes at 3 different scales instead of 2.