**Generalized Focal Loss (GFL)** is a loss function for object detection that combines Quality [Focal Loss](https://paperswithcode.com/method/focal-loss) and Distribution Focal Loss into a general form.

Generalized Focal Loss

Generalized Focal Loss: Learning Qualified and Distributed Bounding Boxes for Dense Object Detection

**Visual-Spatial-Graph Network** (VSGNet) is a network for human-object interaction detection. It extracts visual features from the image representing the human-object pair, refines the features with spatial configurations of the pair, and utilizes the structural connections between the pair via graph convolutions.

VSGNet

VSGNet: Spatial Attention Network for Detecting Human Object Interactions Using Graph Convolutions

**DVD-GAN** is a generative adversarial network for video generation built upon the [BigGAN](https://paperswithcode.com/method/biggan) architecture.

DVD-GAN uses two discriminators: a Spatial Discriminator $\mathcal{D}\_{S}$ and a
Temporal Discriminator $\mathcal{D}\_{T}$. $\mathcal{D}\_{S}$ critiques single frame content and structure by randomly sampling $k$ full-resolution frames and judging them individually.  The temporal discriminator $\mathcal{D}\_{T}$ must provide $G$ with the learning signal to generate movement (not evaluated by $\mathcal{D}\_{S}$).

The input to $G$ consists of a Gaussian latent noise $z \sim N\left(0, I\right)$ and a learned linear embedding $e\left(y\right)$ of the desired class $y$. Both inputs are 120-dimensional vectors. $G$ starts by computing an affine transformation of $\left[z; e\left(y\right)\right]$ to a $\left[4, 4, ch\_{0}\right]$-shaped tensor. $\left[z; e\left(y\right)\right]$ is used as the input to all class-[conditional Batch Normalization](https://paperswithcode.com/method/conditional-batch-normalization) layers
throughout $G$. This is then treated as the input (at each frame we would like to generate) to a Convolutional [GRU](https://paperswithcode.com/method/gru).

This RNN is unrolled once per frame. The output of this RNN is processed by two residual blocks. The time dimension is combined with the batch dimension here, so each frame proceeds through the blocks independently. The output of these blocks has width and height dimensions which
are doubled (we skip upsampling in the first block). This is repeated a number of times, with the
output of one RNN + residual group fed as the input to the next group, until the output tensors have
the desired spatial dimensions. 

The spatial discriminator $\mathcal{D}\_{S}$ functions almost identically to BigGAN’s discriminator. A score is calculated for each of the uniformly sampled $k$ frames (default $k = 8$) and the $\mathcal{D}\_{S}$ output is the sum over per-frame scores. The temporal discriminator $\mathcal{D}\_{T}$ has a similar architecture, but pre-processes the real or generated video with a $2 \times 2$ average-pooling downsampling function $\phi$. Furthermore, the first two residual blocks of $\mathcal{D}\_{T}$ are 3-D, where every [convolution](https://paperswithcode.com/method/convolution) is replaced with a 3-D convolution with a kernel size of $3 \times 3 \times 3$. The rest of the architecture follows BigGAN.

DVD-GAN

Adversarial Video Generation on Complex Datasets

**Hierarchical Transferability Calibration Network** (HTCN) is an adaptive object detector that hierarchically (local-region/image/instance) calibrates the transferability of feature representations for harmonizing transferability and discriminability. The proposed model consists of three components: (1) Importance Weighted Adversarial Training with input Interpolation (IWAT-I), which strengthens the global discriminability by re-weighting the interpolated image-level features; (2) Context-aware Instance-Level Alignment (CILA) module, which enhances the local discriminability by capturing the complementary effect between the instance-level feature and the global context information for the instance-level feature alignment; (3) local feature masks that calibrate the local transferability to provide semantic guidance for the following discriminative pattern alignment.

HTCN

Harmonizing Transferability and Discriminability for Adapting Object Detectors

**ENet Bottleneck** is an image model block used in the [ENet](https://paperswithcode.com/method/enet) semantic segmentation architecture. Each block consists of three convolutional layers: a 1 × 1 projection that reduces the dimensionality, a main convolutional layer, and a 1 × 1 expansion. We place [Batch Normalization](https://paperswithcode.com/method/batch-normalization) and [PReLU](https://paperswithcode.com/method/prelu) between all convolutions. If the bottleneck is downsampling, a [max pooling](https://paperswithcode.com/method/max-pooling) layer is added to the main branch.
Also, the first 1 × 1 projection is replaced with a 2 × 2 [convolution](https://paperswithcode.com/method/convolution) with stride 2 in both dimensions. We zero pad the activations, to match the number of feature maps.

ENet Bottleneck

ENet: A Deep Neural Network Architecture for Real-Time Semantic Segmentation

**Temporal ROI Align** is an operator for extracting features from other frames' feature maps for current frame proposals by utilizing feature similarity. Considering the features of the same object instance are highly similar among frames in a video, the proposed operator implicitly extracts the most similar ROI features from support frames feature map for target frame proposals based on feature similarity.

Temporal ROIAlign

Temporal RoI Align for Video Object Recognition

**Inception v2** is the second generation of Inception convolutional neural network architectures which notably uses [batch normalization](https://paperswithcode.com/method/batch-normalization). Other changes include dropping [dropout](https://paperswithcode.com/method/dropout) and removing [local response normalization](https://paperswithcode.com/method/local-response-normalization), due to the benefits of batch normalization.

Inception v2

Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift

The goal of **Triplet loss**, in the context of Siamese Networks, is to maximize the joint probability among all score-pairs i.e. the product of all probabilities. By using its negative logarithm, we can get the loss formulation as follows:

$$
L\_{t}\left(\mathcal{V}\_{p}, \mathcal{V}\_{n}\right)=-\frac{1}{M N} \sum\_{i}^{M} \sum\_{j}^{N} \log \operatorname{prob}\left(v p\_{i}, v n\_{j}\right)
$$

where the balance weight $1/MN$ is used to keep the loss with the same scale for different number of instance sets.

Triplet Loss

Triplet Loss in Siamese Network for Object Tracking

**NeuroTactic** is a model for theorem proving which leverages [graph neural networks](https://paperswithcode.com/methods/category/graph-models) to represent the theorem and premises, and applies graph contrastive learning for pre-training. Specifically, premise selection is designed as a pretext task for the graph contrastive learning approach. The learned representations are then used for the downstream task, tactic prediction

NeuroTactic

Graph Contrastive Pre-training for Effective Theorem Reasoning

JDeskew is a novel skew estimation method that extracts the dominant skew angle of the given document image by applying an Adaptive Radial Projection on the 2D Discrete Fourier Magnitude spectrum.

JDeskew

Adaptive Radial Projection on Fourier Magnitude Spectrum for Document Image Skew Estimation

**Path Planning and Motion Control**, or **PPMC RL**, is a training algorithm that teaches path planning and motion control to robots using reinforcement learning in a simulated environment. The focus is on promoting generalization where there are environmental uncertainties such as rough environments like lunar services. The algorithm is coupled with any generic reinforcement learning algorithm to teach robots how to respond to user commands and to travel to designated locations on a single neural network. The algorithm works independently of the robot structure, demonstrating that it works on a wheeled rover in addition to the past results on a quadruped walking robot.

PPMC

PPMC RL Training Algorithm: Rough Terrain Intelligent Robots through Reinforcement Learning

GeniePath is a scalable approach for learning adaptive receptive fields of neural networks defined on permutation invariant graph data. In GeniePath, we propose an adaptive path layer consists of two complementary functions designed for breadth and depth exploration respectively, where the former learns the importance of different sized neighborhoods, while the latter extracts and filters signals aggregated from neighbors of different hops away.

Description and image from: [GeniePath: Graph Neural Networks with Adaptive Receptive Paths](https://arxiv.org/pdf/1802.00910.pdf)

GeniePath

GeniePath: Graph Neural Networks with Adaptive Receptive Paths

The ARMA GNN layer implements a rational graph filter with a recursive approximation.

ARMA

Graph Neural Networks with convolutional ARMA filters

**RoBERTa** is an extension of [BERT](https://paperswithcode.com/method/bert) with changes to the pretraining procedure. The modifications include: 

- training the model longer, with bigger batches, over more data
- removing the next sentence prediction objective
- training on longer sequences
- dynamically changing the masking pattern applied to the training data. The authors also collect a large new dataset ($\text{CC-News}$) of comparable size to other privately used datasets, to better control for training set size effects

RoBERTa

RoBERTa: A Robustly Optimized BERT Pretraining Approach

**Corner Pooling** is a pooling technique for object detection that seeks to better localize corners by encoding explicit prior knowledge. Suppose we want to determine if a pixel at location $\left(i, j\right)$ is a top-left corner. Let $f\_{t}$ and $f\_{l}$ be the feature maps that are the inputs to the top-left corner pooling layer, and let $f\_{t\_{ij}}$ and $f\_{l\_{ij}}$ be the vectors at location $\left(i, j\right)$ in $f\_{t}$ and $f\_{l}$ respectively. With $H \times W$ feature maps, the corner pooling layer first max-pools all feature vectors between $\left(i, j\right)$ and $\left(i, H\right)$ in $f\_{t}$ to a feature vector $t\_{ij}$ , and max-pools all feature vectors between $\left(i, j\right)$ and $\left(W, j\right)$ in $f\_{l}$ to a feature vector $l\_{ij}$. Finally, it adds $t\_{ij}$ and $l\_{ij}$ together.

Corner Pooling

CornerNet: Detecting Objects as Paired Keypoints

**VoVNet** is a convolutional neural network that seeks to make [DenseNet](https://paperswithcode.com/method/densenet) more efficient by concatenating all features only once in the last feature map, which makes input size constant and enables enlarging new output channel. In the Figure to the right, $F$ represents a [convolution](https://paperswithcode.com/method/convolution) layer and $\otimes$ indicates concatenation.

VoVNet

An Energy and GPU-Computation Efficient Backbone Network for Real-Time Object Detection

**Aging Evolution**, or **Regularized Evolution**, is an evolutionary algorithm for [neural architecture search](https://paperswithcode.com/method/neural-architecture-search). Whereas in tournament selection, the best architectures are kept, in aging evolution we associate each genotype with an age, and bias the tournament selection to choose
the younger genotypes. In the context of architecture search, aging evolution allows us to explore the search space more, instead of zooming in on good models too early, as non-aging evolution would.

Aging Evolution

Regularized Evolution for Image Classifier Architecture Search

**KnowPrompt** is a prompt-tuning approach for relational understanding. It injects entity and relation knowledge into prompt construction with learnable virtual template words as well as answer words and synergistically optimize their representation with knowledge constraints. To be specific, TYPED MARKER is utilized around entities initialized with aggregated entity-type embeddings as learnable virtual template words to inject entity type knowledge. The average embeddings of each token are leveraged in relation labels as virtual answer words to inject relation knowledge. Since there exist implicit structural constraints among entities and relations, and virtual words should be consistent with the surrounding contexts, synergistic optimization is introduced to obtain optimized virtual templates and answer words. Concretely, a context-aware prompt calibration method is used with implicit structural constraints to inject structural knowledge implications among relational triples and associate prompt embeddings with each other.

Viet-Anh on Software

AI Glossary

Tech Terminology

Generalized Focal Loss

Visual-Spatial-Graph Network