Image Classification

STAC is a semi-supervised framework for visual object detection along with a data augmentation strategy. STAC deploys highly confident pseudo labels of localized objects from an unlabeled image and updates the model by enforcing consistency via strong augmentations. We generate pseudo labels (i.e., bounding boxes and their class labels) for unlabeled data using test-time inference, including NMS , of the teacher model trained with labeled data. We then compute unsupervised loss with respect to pseudo labels whose confidence scores are above a threshold $\tau$ . The strong augmentations are applied for augmentation consistency during the model training. Target boxes are augmented when global geometric transformations are used.

Conditional Positional Encoding, or CPE, is a type of positional encoding for vision transformers. Unlike previous fixed or learnable positional encodings, which are predefined and independent of input tokens, CPE is dynamically generated and conditioned on the local neighborhood of the input tokens. As a result, CPE aims to generalize to the input sequences that are longer than what the model has ever seen during training. CPE can also keep the desired translation-invariance in the image classification task. CPE can be implemented with a Position Encoding Generator (PEG) and incorporated into the current Transformer framework.

FIERCE is an entropic regularization on the feature space

Fast AutoAugment is an image data augmentation algorithm that finds effective augmentation policies via a search strategy based on density matching, motivated by Bayesian DA. The strategy is to improve the generalization performance of a given network by learning the augmentation policies which treat augmented data as missing data points of training data. However, different from Bayesian DA, the proposed method recovers those missing data points by the exploitation-and-exploration of a family of inference-time augmentations via Bayesian optimization in the policy search phase. This is realized by using an efficient density matching algorithm that does not require any back-propagation for network training for each policy evaluation.

GPipe is a distributed model parallel method for neural networks. With GPipe, each model can be specified as a sequence of layers, and consecutive groups of layers can be partitioned into cells. Each cell is then placed on a separate accelerator. Based on this partitioned setup, batch splitting is applied. A mini-batch of training examples is split into smaller micro-batches, then the execution of each set of micro-batches is pipelined over cells. Synchronous mini-batch gradient descent is applied for training, where gradients are accumulated across all micro-batches in a mini-batch and applied at the end of a mini-batch.

A Scale Aggregation Block concatenates feature maps at a wide range of scales. Feature maps for each scale are generated by a stack of downsampling, convolution and upsampling operations. The proposed scale aggregation block is a standard computational module which readily replaces any given transformation $\mathbf{Y}=\mathbf{T}(\mathbf{X})$, where $\mathbf{X}\in \mathbb{R}^{H\times W\times C}$, $\mathbf{Y}\in \mathbb{R}^{H\times W\times C_o}$ with $C$ and $C_o$ being the input and output channel number respectively. $\mathbf{T}$ is any operator such as a convolution layer or a series of convolution layers. Assume we have $L$ scales. Each scale $l$ is generated by sequentially conducting a downsampling $\mathbf{D}_l$, a transformation $\mathbf{T}_l$ and an unsampling operator $\mathbf{U}_l$: $$ \mathbf{X}^{'}_l=\mathbf{D}_l(\mathbf{X}), \label{eq:eq_d} $$ $$ \mathbf{Y}^{'}_l=\mathbf{T}_l(\mathbf{X}^{'}_l), \label{eq:eq_tl} $$ $$ \mathbf{Y}_l=\mathbf{U}_l(\mathbf{Y}^{'}_l), \label{eq:eq_u} $$ where $\mathbf{X}^{'}_l\in \mathbb{R}^{H_l\times W_l\times C}$, $\mathbf{Y}^{'}_l\in \mathbb{R}^{H_l\times W_l\times C_l}$, and $\mathbf{Y}_l\in \mathbb{R}^{H\times W\times C_l}$. Notably, $\mathbf{T}_l$ has the similar structure as $\mathbf{T}$. We can concatenate all $L$ scales together, getting $$ \mathbf{Y}^{'}=\Vert^L_1\mathbf{U}_l(\mathbf{T}_l(\mathbf{D}_l(\mathbf{X}))), \label{eq:eq_all} $$ where $\Vert$ indicates concatenating feature maps along the channel dimension, and $\mathbf{Y}^{'} \in \mathbb{R}^{H\times W\times \sum^L_1 C_l}$ is the final output feature maps of the scale aggregation block. In the reference implementation, the downsampling $\mathbf{D}_l$ with factor $s$ is implemented by a max pool layer with $s\times s$ kernel size and $s$ stride. The upsampling $\mathbf{U}_l$ is implemented by resizing with the nearest neighbor interpolation.

ScaleNet, or a Scale Aggregation Network, is a type of convolutional neural network which learns a neuron allocation for aggregating multi-scale information in different building blocks of a deep network. The most informative output neurons in each block are preserved while others are discarded, and thus neurons for multiple scales are competitively and adaptively allocated. The scale aggregation (SA) block concatenates feature maps at a wide range of scales. Feature maps for each scale are generated by a stack of downsampling, convolution and upsampling operations.

gMLP is an MLP-based alternative to Transformers without self-attention, which simply consists of channel projections and spatial projections with static parameterization. It is built out of basic MLP layers with gating. The model consists of a stack of $L$ blocks with identical size and structure. Let $X \in \mathbb{R}^{n \times d}$ be the token representations with sequence length $n$ and dimension $d$. Each block is defined as: $$ Z=\sigma(X U), \quad \tilde{Z}=s(Z), \quad Y=\tilde{Z} V $$ where $\sigma$ is an activation function such as GeLU. $U$ and $V$ define linear projections along the channel dimension - the same as those in the FFNs of Transformers (e.g., their shapes are $768 \times 3072$ and $3072 \times 768$ for $\text{BERT}_{\text {base }}$). A key ingredient is $s(\cdot)$, a layer which captures spatial interactions. When $s$ is an identity mapping, the above transformation degenerates to a regular FFN, where individual tokens are processed independently without any cross-token communication. One of the major focuses is therefore to design a good $s$ capable of capturing complex spatial interactions across tokens. This leads to the use of a Spatial Gating Unit which involves a modified linear gating. The overall block layout is inspired by inverted bottlenecks, which define $s(\cdot)$ as a spatial depthwise convolution. Note, unlike Transformers, gMLP does not require position embeddings because such information will be captured in $s(\cdot)$.

AdvProp is an adversarial training scheme which treats adversarial examples as additional examples, to prevent overfitting. Key to the method is the usage of a separate auxiliary batch norm for adversarial examples, as they have different underlying distributions to normal examples.

Dilated convolution with learnable spacings (DCLS) is a type of convolution that allows the spacings between the non-zero elements of the kernel to be learned during training. This makes it possible to increase the receptive field of the convolution without increasing the number of parameters, which can improve the performance of the network on tasks that require long-range dependencies. A dilated convolution is a type of convolution that allows the kernel to be skipped over some of the input features. This is done by inserting zeros between the non-zero elements of the kernel. The effect of this is to increase the receptive field of the convolution without increasing the number of parameters. DCLS takes this idea one step further by allowing the spacings between the non-zero elements of the kernel to be learned during training. This means that the network can learn to skip over different input features depending on the task at hand. This can be particularly helpful for tasks that require long-range dependencies, such as image segmentation and object detection. DCLS has been shown to be effective for a variety of tasks, including image classification, object detection, and semantic segmentation. It is a promising new technique that has the potential to improve the performance of convolutional neural networks on a variety of tasks.

A Fractal Block is an image model block that utilizes an expansion rule that yields a structural layout of truncated fractals. For the base case where $f\_{1}\left(z\right) = \text{conv}\left(z\right)$ is a convolutional layer, we then have recursive fractals of the form: $$ f\_{C+1}\left(z\right) = \left[\left(f\_{C}\circ{f\_{C}}\right)\left(z\right)\right] \oplus \left[\text{conv}\left(z\right)\right]$$ Where $C$ is the number of columns. For the join layer (green in Figure), we use the element-wise mean rather than concatenation or addition.

FractalNet is a type of convolutional neural network that eschews residual connections in favour of a "fractal" design. They involve repeated application of a simple expansion rule to generate deep networks whose structural layouts are precisely truncated fractals. These networks contain interacting subpaths of different lengths, but do not include any pass-through or residual connections; every internal signal is transformed by a filter and nonlinearity before being seen by subsequent layers.

NT-ASGD, or Non-monotonically Triggered ASGD, is an averaged stochastic gradient descent technique. In regular ASGD, we take steps identical to regular SGD but instead of returning the last iterate as the solution, we return $\frac{1}{\left(K-T+1\right)}\sum^{T}\_{i=T}w\_{i}$, where $K$ is the total number of iterations and $T < K$ is a user-specified averaging trigger. NT-ASGD has a non-monotonic criterion that conservatively triggers the averaging when the validation metric fails to improve for multiple cycles. Given that the choice of triggering is irreversible, this conservatism ensures that the randomness of training does not play a major role in the decision.

SimpleNet is a convolutional neural network with 13 layers. The network employs a homogeneous design utilizing 3 × 3 kernels for convolutional layer and 2 × 2 kernels for pooling operations. The only layers which do not use 3 × 3 kernels are 11th and 12th layers, these layers, utilize 1 × 1 convolutional kernels. Feature-map down-sampling is carried out using nonoverlaping 2 × 2 max-pooling. In order to cope with the problem of vanishing gradient and also over-fitting, SimpleNet also uses batch-normalization with moving average fraction of 0.95 before any ReLU non-linearity.

VirText, or Visual representations from Textual annotations is a pretraining approach using semantically dense captions to learn visual representations. First a ConvNet and Transformer are jointly trained from scratch to generate natural language captions for images. Then, the learned features are transferred to downstream visual recognition tasks.

ZFNet is a classic convolutional neural network. The design was motivated by visualizing intermediate feature layers and the operation of the classifier. Compared to AlexNet, the filter sizes are reduced and the stride of the convolutions are reduced.

modReLU is an activation that is a modification of a ReLU. It is a pointwise nonlinearity, $\sigma\_{modReLU}\left(z\right) : C \rightarrow C$, which affects only the absolute value of a complex number, defined as: $$ \sigma\_{modReLU}\left(z\right) = \left(|z| + b\right)\frac{z}{|z|} \text{ if } |z| + b \geq 0 $$ $$ \sigma\_{modReLU}\left(z\right) = 0 \text{ if } |z| + b \leq 0 $$ where $b \in \mathbb{R}$ is a bias parameter of the nonlinearity. For a $n\_{h}$ dimensional hidden space we learn $n\_{h}$ nonlinearity bias parameters, one per dimension.

In this work, we present a conceptually simple and effective method to train a strong bilingual multimodal representation model. Starting from the pretrained multimodal representation model CLIP released by OpenAI, we switched its text encoder with a pretrained multilingual text encoder XLM-R, and aligned both languages and image representations by a two-stage training schema consisting of teacher learning and contrastive learning. We validate our method through evaluations of a wide range of tasks. We set new state-of-the-art performances on a bunch of tasks including ImageNet-CN, Flicker30k- CN, and COCO-CN. Further, we obtain very close performances with CLIP on almost all tasks, suggesting that one can simply alter the text encoder in CLIP for extended capabilities such as multilingual understanding. Our models and code are available at https://github.com/FlagAI-Open/FlagAI.

CondConv, or Conditionally Parameterized Convolutions, are a type of convolution which learn specialized convolutional kernels for each example. In particular, we parameterize the convolutional kernels in a CondConv layer as a linear combination of $n$ experts $(\alpha_1 W_1 + \ldots + \alpha_n W_n) * x$, where $\alpha_1, \ldots, \alpha_n$ are functions of the input learned through gradient descent. To efficiently increase the capacity of a CondConv layer, developers can increase the number of experts. This can be more computationally efficient than increasing the size of the convolutional kernel itself, because the convolutional kernel is applied at many different positions within the input, while the experts are combined only once per input.

Contrastive Multiview Coding (CMC) is a self-supervised learning approach, based on CPC, that learns representations that capture information shared between multiple sensory views. The core idea is to set an anchor view and the sample positive and negative data points from the other view and maximise agreement between positive pairs in learning from two views. Contrastive learning is used to build the embedding.

DetNet is a backbone convolutional neural network for object detection. Different from traditional pre-trained models for ImageNet classification, DetNet maintains the spatial resolution of the features even though extra stages are included. DetNet attempts to stay efficient by employing a low complexity dilated bottleneck structure.

An ECA block has similar formulation to an SE block including a squeeze module for aggregating global spatial information and an efficient excitation module for modeling cross-channel interaction. Instead of indirect correspondence, an ECA block only considers direct interaction between each channel and its k-nearest neighbors to control model complexity. Overall, the formulation of an ECA block is: \begin{align} s = F_\text{eca}(X, \theta) & = \sigma (\text{Conv1D}(\text{GAP}(X))) \end{align} \begin{align} Y & = s X \end{align} where $\text{Conv1D}(\cdot)$ denotes 1D convolution with a kernel of shape $k$ across the channel domain, to model local cross-channel interaction. The parameter $k$ decides the coverage of interaction, and in ECA the kernel size $k$ is adaptively determined from the channel dimensionality $C$ instead of by manual tuning, using cross-validation: \begin{equation} k = \psi(C) = \left | \frac{\log_2(C)}{\gamma}+\frac{b}{\gamma}\right |\text{odd} \end{equation} where $\gamma$ and $b$ are hyperparameters. $|x|\text{odd}$ indicates the nearest odd function of $x$. Compared to SENet, ECANet has an improved excitation module, and provides an efficient and effective block which can readily be incorporated into various CNNs.

Recently, dense connections have attracted substantial attention in computer vision because they facilitate gradient flow and implicit deep supervision during training. Particularly, DenseNet that connects each layer to every other layer in a feed-forward fashion and has shown impressive performances in natural image classification tasks. We propose HyperDenseNet, a 3-D fully convolutional neural network that extends the definition of dense connectivity to multi-modal segmentation problems. Each imaging modality has a path, and dense connections occur not only between the pairs of layers within the same path but also between those across different paths. This contrasts with the existing multi-modal CNN approaches, in which modeling several modalities relies entirely on a single joint layer (or level of abstraction) for fusion, typically either at the input or at the output of the network. Therefore, the proposed network has total freedom to learn more complex combinations between the modalities, within and in-between all the levels of abstraction, which increases significantly the learning representation. We report extensive evaluations over two different and highly competitive multi-modal brain tissue segmentation challenges, iSEG 2017 and MRBrainS 2013, with the former focusing on six month infant data and the latter on adult images. HyperDenseNet yielded significant improvements over many state-of-the-art segmentation networks, ranking at the top on both benchmarks. We further provide a comprehensive experimental analysis of features re-use, which confirms the importance of hyper-dense connections in multi-modal representation learning.

Involution is an atomic operation for deep neural networks that inverts the design principles of convolution. Involution kernels are distinct in the spatial extent but shared across channels. If involution kernels are parameterized as fixed-sized matrices like convolution kernels and updated using the back-propagation algorithm, the learned involution kernels are impeded from transferring between input images with variable resolutions. The authors argue for two benefits of involution over convolution: (i) involution can summarize the context in a wider spatial arrangement, thus overcome the difficulty of modeling long-range interactions well; (ii) involution can adaptively allocate the weights over different positions, so as to prioritize the most informative visual elements in the spatial domain.

Meta Pseudo Labels is a semi-supervised learning method that uses a teacher network to generate pseudo labels on unlabeled data to teach a student network. The teacher receives feedback from the student to inform the teacher to generate better pseudo labels. This feedback signal is used as a reward to train the teacher throughout the course of the student’s learning.

Minibatch Discrimination is a discriminative technique for generative adversarial networks where we discriminate between whole minibatches of samples rather than between individual samples. This is intended to avoid collapse of the generator.

MobileViTv2 is a vision transformer that is tuned to mobile device. MobileViTv2 introduced a separable self-attention method to reduce cost than MobileViT

Dynamic Sparse Training method based on using the gradient norm to select which weights to remove/add in each mask update.

SpineNet is a convolutional neural network backbone with scale-permuted intermediate features and cross-scale connections that is learned on an object detection task by Neural Architecture Search.

Lightweight or mobile neural networks used for real-time computer vision tasks contain fewer parameters than normal networks, which lead to a constrained performance. In this work, we proposed a novel activation function named Tanh Exponential Activation Function (TanhExp) which can improve the performance for these networks on image classification task significantly. The definition of TanhExp is $f(x) = x tanh(e^x)$. We demonstrate the simplicity, efficiency, and robustness of TanhExp on various datasets and network models and TanhExp outperforms its counterparts in both convergence speed and accuracy. Its behaviour also remains stable even with noise added and dataset altered. We show that without increasing the size of the network, the capacity of lightweight neural networks can be enhanced by TanhExp with only a few training epochs and no extra parameters added.

Kernel density matrices provide a simpler yet effective mechanism for representing joint probability distributions of both continuous and discrete random variables. This abstraction allows the construction of differentiable models for density estimation, inference, and sampling, and enables their integration into end-to-end deep neural models.

Anti-Alias Downsampling (AA) aims to improve the shift-equivariance of deep networks. Max-pooling is inherently composed of two operations. The first operation is to densely evaluate the max operator and second operation is naive subsampling. AA is proposed as a low-pass filter between them to achieve practical anti-aliasing in any existing strided layer such as strided convolution. The smoothing factor can be adjusted by changing the blur kernel filter size, where a larger filter size results in increased blur.

BigBiGAN is a type of BiGAN with a BigGAN image generator. The authors initially used ResNet as a baseline for the encoder $\mathcal{E}$ followed by a 4-layer MLP with skip connections, but they experimented with RevNets and found they outperformed with increased network width, so opted for this type of encoder for the final architecture.

The focal self-attention is built to make Transformer layers scalable to high-resolution inputs. Instead of attending all tokens at fine-grain, the approach attends the fine-grain tokens only locally, but the summarized ones globally. As such, it can cover as many regions as standard self-attention but with much less cost. An image is first partitioned into patches, resulting in visual tokens. Then a patch embedding layer, consisting of a convolutional layer with filter and stride of same size, to project the patches into hidden features. This spatial feature map in then passed to four stages of focal Transformer blocks. Each focal Transformer block consists of $N_i$ focal Transformer layers. Patch embedding layers are used in between to reduce spatial size of feature map by factor 2, while feature dimension increased by 2.