S. Balasubramanian

Darshan Gera, Badveeti Naveen Siva Kumar, Bobbili Veerendra Raj Kumar et al.

Automatic affect recognition has applications in many areas such as education, gaming, software development, automotives, medical care, etc. but it is non trivial task to achieve appreciable performance on in-the-wild data sets. In-the-wild data sets though represent real-world scenarios better than synthetic data sets, the former ones suffer from the problem of incomplete labels. Inspired by semi-supervised learning, in this paper, we introduce our submission to the Multi-Task-Learning Challenge at the 4th Affective Behavior Analysis in-the-wild (ABAW) 2022 Competition. The three tasks that are considered in this challenge are valence-arousal(VA) estimation, classification of expressions into 6 basic (anger, disgust, fear, happiness, sadness, surprise), neutral, and the 'other' category and 12 action units(AU) numbered AU-{1,2,4,6,7,10,12,15,23,24,25,26}. Our method Semi-supervised Multi-task Facial Affect Recognition titled SS-MFAR uses a deep residual network with task specific classifiers for each of the tasks along with adaptive thresholds for each expression class and semi-supervised learning for the incomplete labels. Source code is available at https://github.com/1980x/ABAW2022DMACS.

Darshan Gera, Naveen Siva Kumar Badveeti, Bobbili Veerendra Raj Kumar et al.

The real-world facial expression recognition (FER) datasets suffer from noisy annotations due to crowd-sourcing, ambiguity in expressions, the subjectivity of annotators and inter-class similarity. However, the recent deep networks have strong capacity to memorize the noisy annotations leading to corrupted feature embedding and poor generalization. To handle noisy annotations, we propose a dynamic FER learning framework (DNFER) in which clean samples are selected based on dynamic class specific threshold during training. Specifically, DNFER is based on supervised training using selected clean samples and unsupervised consistent training using all the samples. During training, the mean posterior class probabilities of each mini-batch is used as dynamic class-specific threshold to select the clean samples for supervised training. This threshold is independent of noise rate and does not need any clean data unlike other methods. In addition, to learn from all samples, the posterior distributions between weakly-augmented image and strongly-augmented image are aligned using an unsupervised consistency loss. We demonstrate the robustness of DNFER on both synthetic as well as on real noisy annotated FER datasets like RAFDB, FERPlus, SFEW and AffectNet.

Darshan Gera, Badveeti Naveen Siva Kumar, Bobbili Veerendra Raj Kumar et al.

The fifth Affective Behavior Analysis in-the-wild (ABAW) competition has multiple challenges such as Valence-Arousal Estimation Challenge, Expression Classification Challenge, Action Unit Detection Challenge, Emotional Reaction Intensity Estimation Challenge. In this paper we have dealt only expression classification challenge using multiple approaches such as fully supervised, semi-supervised and noisy label approach. Our approach using noise aware model has performed better than baseline model by 10.46% and semi supervised model has performed better than baseline model by 9.38% and the fully supervised model has performed better than the baseline by 9.34%

Sai Raam Venkataraman, S. Balasubramanian, R. Raghunatha Sarma

Geometric transformations of the training data as well as the test data present challenges to the use of deep neural networks to vision-based learning tasks. In order to address this issue, we present a deep neural network model that exhibits the desirable property of transformation-robustness. Our model, termed RobustCaps, uses group-equivariant convolutions in an improved capsule network model. RobustCaps uses a global context-normalised procedure in its routing algorithm to learn transformation-invariant part-whole relationships within image data. This learning of such relationships allows our model to outperform both capsule and convolutional neural network baselines on transformation-robust classification tasks. Specifically, RobustCaps achieves state-of-the-art accuracies on CIFAR-10, FashionMNIST, and CIFAR-100 when the images in these datasets are subjected to train and test-time rotations and translations.

Sai Raam Venkataraman, S. Balasubramanian, R. Raghunatha Sarma

Transformation-robustness is an important feature for machine learning models that perform image classification. Many methods aim to bestow this property to models by the use of data augmentation strategies, while more formal guarantees are obtained via the use of equivariant models. We recognise that compositional, or part-whole structure is also an important aspect of images that has to be considered for building transformation-robust models. Thus, we propose a capsule network model that is, at once, equivariant and compositionality-aware. Equivariance of our capsule network model comes from the use of equivariant convolutions in a carefully-chosen novel architecture. The awareness of compositionality comes from the use of our proposed novel, iterative, graph-based routing algorithm, termed Iterative collaborative routing (ICR). ICR, the core of our contribution, weights the predictions made for capsules based on an iteratively averaged score of the degree-centralities of its nearest neighbours. Experiments on transformed image classification on FashionMNIST, CIFAR-10, and CIFAR-100 show that our model that uses ICR outperforms convolutional and capsule baselines to achieve state-of-the-art performance.

Sai Raam Venkatraman, Rishi Rao, S. Balasubramanian et al.

Visual Question Answering (VQA) needs a means of evaluating the strengths and weaknesses of models. One aspect of such an evaluation is the evaluation of compositional generalisation, or the ability of a model to answer well on scenes whose scene-setups are different from the training set. Therefore, for this purpose, we need datasets whose train and test sets differ significantly in composition. In this work, we present several quantitative measures of compositional separation and find that popular datasets for VQA are not good evaluators. To solve this, we present Uncommon Objects in Unseen Configurations (UOUC), a synthetic dataset for VQA. UOUC is at once fairly complex while also being well-separated, compositionally. The object-class of UOUC consists of 380 clasess taken from 528 characters from the Dungeons and Dragons game. The train set of UOUC consists of 200,000 scenes; whereas the test set consists of 30,000 scenes. In order to study compositional generalisation, simple reasoning and memorisation, each scene of UOUC is annotated with up to 10 novel questions. These deal with spatial relationships, hypothetical changes to scenes, counting, comparison, memorisation and memory-based reasoning. In total, UOUC presents over 2 million questions. UOUC also finds itself as a strong challenge to well-performing models for VQA. Our evaluation of recent models for VQA shows poor compositional generalisation, and comparatively lower ability towards simple reasoning. These results suggest that UOUC could lead to advances in research by being a strong benchmark for VQA.

Darshan Gera, S. Balasubramanian

Presence of noise in the labels of large scale facial expression datasets has been a key challenge towards Facial Expression Recognition (FER) in the wild. During early learning stage, deep networks fit on clean data. Then, eventually, they start overfitting on noisy labels due to their memorization ability, which limits FER performance. This work proposes an effective training strategy in the presence of noisy labels, called as Consensual Collaborative Training (CCT) framework. CCT co-trains three networks jointly using a convex combination of supervision loss and consistency loss, without making any assumption about the noise distribution. A dynamic transition mechanism is used to move from supervision loss in early learning to consistency loss for consensus of predictions among networks in the later stage. Inference is done using a single network based on a simple knowledge distillation scheme. Effectiveness of the proposed framework is demonstrated on synthetic as well as real noisy FER datasets. In addition, a large test subset of around 5K images is annotated from the FEC dataset using crowd wisdom of 16 different annotators and reliable labels are inferred. CCT is also validated on it. State-of-the-art performance is reported on the benchmark FER datasets RAFDB (90.84%) FERPlus (89.99%) and AffectNet (66%). Our codes are available at https://github.com/1980x/CCT.

Darshan Gera, S Balasubramanian

Facial expression recognition (FER) in the wild is crucial for building reliable human-computer interactive systems. However, annotations of large scale datasets in FER has been a key challenge as these datasets suffer from noise due to various factors like crowd sourcing, subjectivity of annotators, poor quality of images, automatic labelling based on key word search etc. Such noisy annotations impede the performance of FER due to the memorization ability of deep networks. During early learning stage, deep networks fit on clean data. Then, eventually, they start overfitting on noisy labels due to their memorization ability, which limits FER performance. This report presents Consensual Collaborative Training (CCT) framework used in our submission to expression recognition track of the Affective Behaviour Analysis in-the-wild (ABAW) 2021 competition. CCT co-trains three networks jointly using a convex combination of supervision loss and consistency loss, without making any assumption about the noise distribution. A dynamic transition mechanism is used to move from supervision loss in early learning to consistency loss for consensus of predictions among networks in the later stage. Co-training reduces overall error, and consistency loss prevents overfitting to noisy samples. The performance of the model is validated on challenging Aff-Wild2 dataset for categorical expression classification. Our code is made publicly available at https://github.com/1980x/ABAW2021DMACS.

Darshan Gera, S Balasubramanian

A recent trend to recognize facial expressions in the real-world scenario is to deploy attention based convolutional neural networks (CNNs) locally to signify the importance of facial regions and, combine it with global facial features and/or other complementary context information for performance gain. However, in the presence of occlusions and pose variations, different channels respond differently, and further that the response intensity of a channel differ across spatial locations. Also, modern facial expression recognition(FER) architectures rely on external sources like landmark detectors for defining attention. Failure of landmark detector will have a cascading effect on FER. Additionally, there is no emphasis laid on the relevance of features that are input to compute complementary context information. Leveraging on the aforementioned observations, an end-to-end architecture for FER is proposed in this work that obtains both local and global attention per channel per spatial location through a novel spatio-channel attention net (SCAN), without seeking any information from the landmark detectors. SCAN is complemented by a complementary context information (CCI) branch. Further, using efficient channel attention (ECA), the relevance of features input to CCI is also attended to. The representation learnt by the proposed architecture is robust to occlusions and pose variations. Robustness and superior performance of the proposed model is demonstrated on both in-lab and in-the-wild datasets (AffectNet, FERPlus, RAF-DB, FED-RO, SFEW, CK+, Oulu-CASIA and JAFFE) along with a couple of constructed face mask datasets resembling masked faces in COVID-19 scenario. Codes are publicly available at https://github.com/1980x/SCAN-CCI-FER

Darshan Gera, Badveeti Naveen Siva Kumar, Bobbili Veerendra Raj Kumar et al.

The hindering problem in facial expression recognition (FER) is the presence of inaccurate annotations referred to as noisy annotations in the datasets. These noisy annotations are present in the datasets inherently because the labeling is subjective to the annotator, clarity of the image, etc. Recent works use sample selection methods to solve this noisy annotation problem in FER. In our work, we use a dynamic adaptive threshold to separate confident samples from non-confident ones so that our learning won't be hampered due to non-confident samples. Instead of discarding the non-confident samples, we impose consistency in the negative classes of those non-confident samples to guide the model to learn better in the positive class. Since FER datasets usually come with 7 or 8 classes, we can correctly guess a negative class by 85% probability even by choosing randomly. By learning "which class a sample doesn't belong to", the model can learn "which class it belongs to" in a better manner. We demonstrate proposed framework's effectiveness using quantitative as well as qualitative results. Our method performs better than the baseline by a margin of 4% to 28% on RAFDB and 3.3% to 31.4% on FERPlus for various levels of synthetic noisy labels in the aforementioned datasets.

Darshan Gera, S. Balasubramanian

Since the renaissance of deep learning (DL), facial expression recognition (FER) has received a lot of interest, with continual improvement in the performance. Hand-in-hand with performance, new challenges have come up. Modern FER systems deal with face images captured under uncontrolled conditions (also called in-the-wild scenario) including occlusions and pose variations. They successfully handle such conditions using deep networks that come with various components like transfer learning, attention mechanism and local-global context extractor. However, these deep networks are highly complex with large number of parameters, making them unfit to be deployed in real scenarios. Is it possible to build a light-weight network that can still show significantly good performance on FER under in-the-wild scenario? In this work, we methodically build such a network and call it as Imponderous Net. We leverage on the aforementioned components of deep networks for FER, and analyse, carefully choose and fit them to arrive at Imponderous Net. Our Imponderous Net is a low calorie net with only 1.45M parameters, which is almost 50x less than that of a state-of-the-art (SOTA) architecture. Further, during inference, it can process at the real time rate of 40 frames per second (fps) in an intel-i7 cpu. Though it is low calorie, it is still power packed in its performance, overpowering other light-weight architectures and even few high capacity architectures. Specifically, Imponderous Net reports 87.09\%, 88.17\% and 62.06\% accuracies on in-the-wild datasets RAFDB, FERPlus and AffectNet respectively. It also exhibits superior robustness under occlusions and pose variations in comparison to other light-weight architectures from the literature.

Sai Raam Venkatraman, Ankit Anand, S. Balasubramanian et al.

A formal description of the compositionality of neural networks is associated directly with the formal grammar-structure of the objects it seeks to represent. This formal grammar-structure specifies the kind of components that make up an object, and also the configurations they are allowed to be in. In other words, objects can be described as a parse-tree of its components -- a structure that can be seen as a candidate for building connection-patterns among neurons in neural networks. We present a formal grammar description of convolutional neural networks and capsule networks that shows how capsule networks can enforce such parse-tree structures, while CNNs do not. Specifically, we show that the entropy of routing coefficients in the dynamic routing algorithm controls this ability. Thus, we introduce the entropy of routing weights as a loss function for better compositionality among capsules. We show by experiments, on data with a compositional structure, that the use of this loss enables capsule networks to better detect changes in compositionality. Our experiments show that as the entropy of the routing weights increases, the ability to detect changes in compositionality reduces. We see that, without routing, capsule networks perform similar to convolutional neural networks in that both these models perform badly at detecting changes in compositionality. Our results indicate that routing is an important part of capsule networks -- effectively answering recent work that has questioned its necessity. We also, by experiments on SmallNORB, CIFAR-10, and FashionMNIST, show that this loss keeps the accuracy of capsule network models comparable to models that do not use it .

Darshan Gera, S Balasubramanian

Facial expression recognition(FER) in the wild is crucial for building reliable human-computer interactive systems. However, current FER systems fail to perform well under various natural and un-controlled conditions. This report presents attention based framework used in our submission to expression recognition track of the Affective Behaviour Analysis in-the-wild (ABAW) 2020 competition. Spatial-channel attention net(SCAN) is used to extract local and global attentive features without seeking any information from landmark detectors. SCAN is complemented by a complementary context information(CCI) branch which uses efficient channel attention(ECA) to enhance the relevance of features. The performance of the model is validated on challenging Aff-Wild2 dataset for categorical expression classification.

Sairaam Venkatraman, S. Balasubramanian, R. Raghunatha Sarma

Capsule networks are constrained by the parameter-expensive nature of their layers, and the general lack of provable equivariance guarantees. We present a variation of capsule networks that aims to remedy this. We identify that learning all pair-wise part-whole relationships between capsules of successive layers is inefficient. Further, we also realise that the choice of prediction networks and the routing mechanism are both key to equivariance. Based on these, we propose an alternative framework for capsule networks that learns to projectively encode the manifold of pose-variations, termed the space-of-variation (SOV), for every capsule-type of each layer. This is done using a trainable, equivariant function defined over a grid of group-transformations. Thus, the prediction-phase of routing involves projection into the SOV of a deeper capsule using the corresponding function. As a specific instantiation of this idea, and also in order to reap the benefits of increased parameter-sharing, we use type-homogeneous group-equivariant convolutions of shallower capsules in this phase. We also introduce an equivariant routing mechanism based on degree-centrality. We show that this particular instance of our general model is equivariant, and hence preserves the compositional representation of an input under transformations. We conduct several experiments on standard object-classification datasets that showcase the increased transformation-robustness, as well as general performance, of our model to several capsule baselines.

Varshaneya V, S Balasubramanian, Vineeth N Balasubramanian

Sketching is more fundamental to human cognition than speech. Deep Neural Networks (DNNs) have achieved the state-of-the-art in speech-related tasks but have not made significant development in generating stroke-based sketches a.k.a sketches in vector format. Though there are Variational Auto Encoders (VAEs) for generating sketches in vector format, there is no Generative Adversarial Network (GAN) architecture for the same. In this paper, we propose a standalone GAN architecture SkeGAN and a VAE-GAN architecture VASkeGAN, for sketch generation in vector format. SkeGAN is a stochastic policy in Reinforcement Learning (RL), capable of generating both multidimensional continuous and discrete outputs. VASkeGAN hybridizes a VAE and a GAN, in order to couple the efficient representation of data by VAE with the powerful generating capabilities of a GAN, to produce visually appealing sketches. We also propose a new metric called the Ske-score which quantifies the quality of vector sketches. We have validated that SkeGAN and VASkeGAN generate visually appealing sketches by using Human Turing Test and Ske-score.

Varshaneya V, Balasubramanian S, Darshan Gera

One of the ways to train deep neural networks effectively is to use residual connections. Residual connections can be classified as being either identity connections or bridge-connections with a reshaping convolution. Empirical observations on CIFAR-10 and CIFAR-100 datasets using a baseline Resnet model, with bridge-connections removed, have shown a significant reduction in accuracy. This reduction is due to lack of contribution, in the form of feature maps, by the bridge-connections. Hence bridge-connections are vital for Resnet. However, all feature maps in the bridge-connections are considered to be equally important. In this work, an upgraded architecture "Res-SE-Net" is proposed to further strengthen the contribution from the bridge-connections by quantifying the importance of each feature map and weighting them accordingly using Squeeze-and-Excitation (SE) block. It is demonstrated that Res-SE-Net generalizes much better than Resnet and SE-Resnet on the benchmark CIFAR-10 and CIFAR-100 datasets.