Hierarchical reinforcement learning (HRL) leverages temporal abstraction to efficiently tackle complex long-horizon tasks. However, HRL often collapses because the low-level primitive’s continual updates make earlier sub-goals issued by the high-level policy obsolete, introducing non-stationarity that destabilizes training. We propose CRISP, a curriculum-driven framework that tackles this instability with three key ingredients: (1) primitive-informed parsing (PIP), which adaptively re-labels a handful of expert demonstrations to always generate reachable subgoals by the current low-level primitive; (2) an inverse-reinforcement-learning regularizer that steers the high-level policy toward the expert-induced subgoal distribution and stabilizes learning; and (3) a unified training loop that leverages these components to boost sample efficiency. Across six sparse-reward robotic navigation and manipulation benchmarks, CRISP improves success rates by more than 40% over strong hierarchical and flat baselines and successfully transfers to real-world tasks, demonstrating the promise of curriculum-based HRL for practical scenarios.

Zero-shot, training-free, image-based text-to-video generation is an emerging area that aims to generate videos using existing image-based diffusion models. Current methods in this space require specific architectural changes to image-generation models, which limit their adaptability and scalability. In contrast to such methods, we provide a model-agnostic approach. We use intersections in diffusion trajectories, working only with the latent values. We could not obtain localized frame-wise coherence and diversity using only the intersection of trajectories. Thus, we instead use a grid-based approach. An in-context trained LLM is used to generate coherent frame-wise prompts; another is used to identify differences between frames. Based on these, we obtain a CLIP-based attention mask that controls the timing of switching the prompts for each grid cell. Earlier switching results in higher variance, while later switching results in more coherence. Therefore, Our approach can ensure appropriate control between coherence and variance for the frames. Our approach results in state-of-the-art performance while being more flexible when working with diverse image-generation models. The empirical analysis using quantitative metrics and user studies confirms our model’s superior temporal consistency, visual fidelity and user satisfaction, thus providing a novel way to obtain training-free, image-based text-to-video generation. Further examples and code at https://djagpal02.github.io/EIDT-V/

Gaussian Splatting has changed the game for real-time photo-realistic rendering. One of the most popular applications of Gaussian Splatting is to create animatable avatars, known as Gaussian Avatars. Recent works have pushed the boundaries of quality and rendering efficiency but suffer from two main limitations. Either they require expensive multi-camera rigs to produce avatars with free-viewpoint rendering, or they can be trained with a single camera but only rendered at high quality from this fixed viewpoint. An ideal model would be trained using a short monocular video or image from available hardware, such as a webcam, and rendered from any view. To this end, we propose GASP: Gaussian Avatars with Synthetic Priors. To overcome the limitations of existing datasets, we exploit the pixel-perfect nature of synthetic data to train a Gaussian Avatar prior. By fitting this prior model to a single photo or video and fine-tuning it, we get a high-quality Gaussian Avatar, which supports 360° rendering. Our prior is only required for fitting, not inference, enabling real-time applications. Through our method, we obtain high-quality, animatable Avatars from limited data which can be animated and rendered at 70fps on commercial hardware.

With the popularity of foundational models, parameter efficient fine tuning has become the defacto approach to leverage pretrained models to perform downstream tasks. Taking inspiration from recent advances in large language models, Visual Prompt Tuning, and similar techniques, learn an additional prompt to efficiently finetune a pretrained vision foundational model. However, we observe that such prompting is insufficient for fine-grained visual classification tasks such as medical image classification, where there is large inter-class variance, and small intra-class variance. Hence, in this paper we propose to leverage advanced segmentation capabilities of Segment Anything Model 2 [1] (SAM2) as a visual prompting cue to help visual encoder in the CLIP [2] (Contrastive Language-Image Pretraining) by guiding the attention in CLIP visual encoder to relevant regions in the image. This helps the model to focus on highly discriminative regions, without getting distracted from visually similar background features, an essential requirement in a fewshot, finegrained classification setting. We evaluate our method on diverse medical datasets including X-rays, CT scans, and MRI images, and report an accuracy of (71%, 81%, 86%, 58%) from the proposed approach on (COVID, lung-disease, brain-tumor, breast-cancer) datasets against (66%, 70%, 68%, 29%) from a pretrained CLIP model after fewshot training. The proposed approach also allows to obtain interpretable explanation for the classification performance through the localization obtained using segmentation. For demonstrations and visualizations, please visit https://aadya-arora.github.io/MedFocusClip/

Hierarchical reinforcement learning (HRL) has the potential to solve complex long horizon tasks using temporal abstraction and increased exploration. However, hierarchical agents are difficult to train due to inherent non-stationarity. We present primitive enabled adaptive relabeling (PEAR), a two-phase approach where we first perform adaptive relabeling on a few expert demonstrations to generate efficient subgoal supervision, and then jointly optimize HRL agents by employing reinforcement learning (RL) and imitation learning (IL). We perform theoretical analysis to bound the sub-optimality of our approach and derive a joint optimization framework using RL and IL. Since PEAR utilizes only a few expert demonstrations and considers minimal limiting assumptions on the task structure, it can be easily integrated with typical off-policy \RL algorithms to produce a practical HRL approach. We perform extensive experiments on challenging environments and show that PEAR is able to outperform various hierarchical and non-hierarchical baselines and achieve upto 80% success rates in complex sparse robotic control tasks where other baselines typically fail to show significant progress. We also perform ablations to thoroughly analyze the importance of our various design choices. Finally, we perform real world robotic experiments on complex tasks and demonstrate that PEAR consistently outperforms the baselines.

Machine unlearning is a crucial area of research. It is driven by the need to remove sensitive information from models to safeguard individuals' right to be forgotten under rigorous regulations such as GDPR. In this work, we focus on unlearning within CLIP, a dual vision-language encoder model trained on a massive dataset of image-text pairs using contrastive loss. To achieve forgetting we expand the application of Lipschitz regularization to the multimodal context of CLIP. Specifically, we smooth both visual and textual embeddings associated with the class intended to be forgotten relative to the perturbation introduced to the samples from that class. Additionally, importantly, we remove the necessity for real forgetting data by generating synthetic samples via gradient ascent maximizing the target class. Our forgetting procedure is iterative, where we track accuracy on a synthetic forget set and stop when accuracy falls below a chosen threshold. We employ a selective layers update strategy based on their average absolute gradient value to mitigate over-forgetting. We validate our approach on several standard datasets and provide thorough ablation analysis and comparisons with previous work.

Vision transformers have been applied successfully for image recognition tasks. There have been either multi-headed self-attention based (ViT [12], DelT [54]) simi-lar to the original work in textual models or more re-cently based on spectral layers (Fnet [29], GFNet [46], AFNO [15]). We hypothesize that spectral layers cap-ture high-frequency information such as lines and edges, while attention layers capture token interactions. We inves-tigate this hypothesis through this work and observe that indeed mixing spectral and multi-headed attention layers provides a better transformer architecture. We thus pro-pose the novel Spectformer architecture for vision trans-formers that has initial spectral and deeper multi-headed attention layers. We believe that the resulting representation allows the transformer to capture the feature repre-sentation appropriately and it yields improved performance over other transformer representations. For instance, it im-proves the top-1 accuracy by 2% on ImageNet compared to both GFNet-H and LiT. SpectFormer-H-S reaches 84.25% top-1 accuracy on ImageNet-1 K (state of the art for small version). Further, Spectformer-H-L achieves 85.7% which is the state of the art for the comparable base version of the transformers. We further validated the SpectFormer per-formance in other scenarios such as transfer learning on standard datasets such as ClFAR-10, ClFAR-100, Oxford-lIlT-flower, and Standford Car datasets. We then investigate its use in downstream tasks such as object detection and instance segmentation on the MS-COCO dataset and ob-serve that Spectformer shows consistent performance that is comparable to the best backbones and can be further optimized and improved. The source code is available on this website https://github.com/badripatro/SpectFormers.

This work tackles the important task of understanding out-of-distribution behavior in two prominent types of generative models, i.e., GANs and Diffusion models. Understanding this behavior is crucial in understanding their broader utility and risks as these systems are increasingly deployed in our daily lives. Our first contribution is demonstrating that diffusion spaces outperform GANs' latent spaces in inverting high-quality OOD images. We also provide a theoretical analysis attributing this to the lack of prior holes in diffusion spaces. Our second significant contribution is to provide a theoretical hypothesis that diffusion spaces can be projected onto a bounded hypersphere, enabling image manipulation through geodesic traversal between inverted images. Our analysis shows that different geodesics share common attributes for the same manipulation, which we leverage to perform various image manipulations. We conduct thorough empirical evaluations to support and validate our claims. Finally, our third and final contribution introduces a novel approach to the few-shot sampling for out-of-distribution data by inverting a few images to sample from the cluster formed by the inverted latents. The proposed technique achieves state-of-the-art results for the few-shot generation task in terms of image quality. Our research underscores the promise of diffusion spaces in out-of-distribution imaging and offers avenues for further exploration. Please find more details about our project at \url{http://cvit.iiit.ac.in/research/projects/cvit-projects/diffusionOOD}

In this work, we introduce PIPER: Primitive-Informed Preference-based Hierarchical reinforcement learning via Hindsight Relabeling, a novel approach that leverages preference-based learning to learn a reward model, and subsequently uses this reward model to relabel higher-level replay buffers. Since this reward is unaffected by lower primitive behavior, our relabeling-based approach is able to mitigate non-stationarity, which is common in existing hierarchical approaches, and demonstrates impressive performance across a range of challenging sparse-reward tasks. Since obtaining human feedback is typically impractical, we propose to replace the human-in-the-loop approach with our primitive-in-the-loop approach, which generates feedback using sparse rewards provided by the environment. Moreover, in order to prevent infeasible subgoal prediction and avoid degenerate solutions, we propose primitive-informed regularization that conditions higher-level policies to generate feasible subgoals for lower-level policies. We perform extensive experiments to show that PIPER mitigates non-stationarity in hierarchical reinforcement learning and achieves greater than 50$\%$ success rates in challenging, sparse-reward robotic environments, where most other baselines fail to achieve any significant progress.

Lifelong learning or continual learning is the problem of training an AI agent continuously while also preventing it from forgetting its previously acquired knowledge. Streaming lifelong learning is a challenging setting of lifelong learning with the goal of continuous learning in a dynamic non-stationary environment without forgetting. We introduce a novel approach to lifelong learning, which is streaming (observes each training example only once), requires a single pass over the data, can learn in a class-incremental manner, and can be evaluated on-the-fly (anytime inference). To accomplish these, we propose a novel virtual gradients based approach for continual representation learning which adapts to each new example while also generalizing well on past data to prevent catastrophic forgetting. Our approach also leverages an exponential-moving-average-based semantic memory to further enhance performance. Experiments on diverse datasets with temporally correlated observations demonstrate our method’s efficacy and superior performance over existing methods.

In the task incremental learning problem, deep learning models suffer from catastrophic forgetting of previously seen classes/tasks as they are trained on new classes/tasks. This problem becomes even harder when some of the test classes do not belong to the training class set, i.e., the task incremental generalized zero-shot learning problem. We propose a novel approach to address the task incremental learning problem for both the non zero-shot and zero-shot settings. Our proposed approach, called Rectification-based Knowledge Retention (RKR), applies weight rectifications and affine transformations for adapting the model to any task. During testing, our approach can use the task label information (task-aware) to quickly adapt the network to that task. We also extend our approach to make it task-agnostic so that it can work even when the task label information is not available during testing. Specifically, given a continuum of test data, our approach predicts the task and quickly adapts the network to the predicted task. We experimentally show that our proposed approach achieves state-of-the-art results on several benchmark datasets for both non zero-shot and zero-shot task incremental learning.

Despite rapid advancements in the lifelong learning (LL) research, a large body of research mainly focuses on improving the performance in the existing static continual learning (CL) setups. These methods lack the ability to succeed in a rapidly changing dynamic environment, where an AI agent needs to quickly learn new instances in a ‘single pass' from the non-i.i.d (also possibly temporally contiguous/coherent) data streams without suffering from catastrophic forgetting. For practical applicability, we propose a novel lifelong learning approach, which is streaming, i.e., a single input sample arrives in each time step. Moreover, the proposed approach is single pass, class-incremental, and is subject to be evaluated at any moment. To address this challenging setup and various evaluation protocols, we propose a Bayesian framework, that enables fast parameter update, given a single training example, and enables any-time inference. We additionally propose an implicit regularizer in the form of snap-shot self-distillation, which effectively minimizes the forgetting further. We further propose an effective method that efficiently selects a subset of samples for online memory rehearsal and employs a new replay buffer management scheme that significantly boosts the overall performance. Our empirical evaluations and ablations demonstrate that the proposed method outperforms the prior works by large margins.

In this paper, we introduce a novel approach to address the task of synthesizing speech from silent videos of any in-the-wild speaker solely based on lip movements. The traditional approach of directly generating speech from lip videos faces the challenge of not being able to learn a robust language model from speech alone, resulting in unsatisfactory outcomes. To overcome this issue, we propose incorporating noisy text supervision using a state-of-the-art lip-to-text network that instills language information into our model. The noisy text is generated using a pre-trained lip-to-text model, enabling our approach to work without text annotations during inference. We design a visual text-to-speech network that utilizes the visual stream to generate accurate speech, which is in-sync with the silent input video. We perform extensive experiments and ablation studies, demonstrating our approach's superiority over the current state-of-the-art methods on various benchmark datasets. Further, we demonstrate an essential practical application of our method in assistive technology by generating speech for an ALS patient who has lost the voice but can make mouth movements. Our demo video, code, and additional details can be found at http://cvit.iiit.ac.in/research/projects/cvit-projects/ms-l2s-itw.

This work proposes a novel method to generate realistic talking head videos using audio and visual streams. We animate a source image by transferring head motion from a driving video using a dense motion field generated using learnable keypoints. We improve the quality of lip sync using audio as an additional input, helping the network to attend to the mouth region. We use additional priors using face segmentation and face mesh to improve the structure of the reconstructed faces. Finally, we improve the visual quality of the generations by incorporating a carefully designed identity-aware generator module. The identity-aware generator takes the source image and the warped motion features as input to generate a high-quality output with fine-grained details. Our method produces state-of-the-art results and generalizes well to unseen faces, languages, and voices. We comprehensively evaluate our approach using multiple metrics and outperforming the current techniques both qualitative and quantitatively. Our work opens up several applications, including enabling low bandwidth video calls. We release a demo video and additional information at http://cvit.iiit.ac.in/research/projects/cvit-projects/avfr.

Many people with some form of hearing loss consider lipreading as their primary mode of day-to-day communication. However, finding resources to learn or improve one’s lipreading skills can be challenging. This is further exacerbated in the COVID19 pandemic due to restrictions on direct interactions with peers and speech therapists. Today, online MOOCs platforms like Coursera and Udemy have become the most effective form of training for many types of skill development. However, online lipreading resources are scarce as creating such resources is an extensive process needing months of manual effort to record hired ac-tors. Because of the manual pipeline, such platforms are also limited in vocabulary, supported languages, accents, and speakers and have a high usage cost. In this work, we investigate the possibility of replacing real human talking videos with synthetically generated videos. Synthetic data can easily incorporate larger vocabularies, variations in accent, and even local languages and many speakers. We propose an end-to-end automated pipeline to develop such a platform using state-of-the-art talking head video generator networks, text-to-speech models, and computer vision techniques. We then perform an extensive human evaluation using carefully thought out lipreading exercises to validate the quality of our designed platform against the existing lipreading platforms. Our studies concretely point toward the potential of our approach in developing a large-scale lipreading MOOC platform that can impact millions of people with hearing loss.

Doubles play an indispensable role in the movie industry. They take the place of the actors in dangerous stunt scenes or scenes where the same actor plays multiple characters. The double's face is later replaced with the actor's face and expressions manually using expensive CGI technology, costing millions of dollars and taking months to complete. An automated, inexpensive, and fast way can be to use face-swapping techniques that aim to swap an identity from a source face video (or an image) to a target face video. However, such methods cannot preserve the source expressions of the actor important for the scene's context. To tackle this challenge, we introduce video-to-video (V2V) face-swapping, a novel task of face-swapping that can preserve (1) the identity and expressions of the source (actor) face video and (2) the background and pose of the target (double) video. We propose FaceOff, a V2V face-swapping system that operates by learning a robust blending operation to merge two face videos following the constraints above. It reduces the videos to a quantized latent space and then blends them in the reduced space. FaceOff is trained in a self-supervised manner and robustly tackles the non-trivial challenges of V2V face-swapping. As shown in the experimental section, FaceOff significantly outperforms alternate approaches qualitatively and quantitatively.

Talking-face video generation works have achieved state-of-the-art results in synthesizing videos with lip synchronization. However, most of the previous works deal with low-resolution talking-face videos (up to 256×256 pixels), thus, generating extremely high-resolution videos still remains a challenge. We take a giant leap in this work and propose a novel method to synthesize talking-face videos at resolutions as high as 4K! Our task presents several key challenges: (i) Scaling the existing methods to such high resolutions is resource-constrained, both in terms of compute and the availability of very high-resolution datasets, (ii) The synthesized videos need to be spatially and temporally coherent. The sheer number of pixels that the model needs to generate while maintaining the temporal consistency at the video level makes this task non-trivial and has never been attempted before in literature. To address these issues, we propose to train the lip-sync generator in a compact Vector Quantized (VQ) space for the first time. Our core idea to encode the faces in a compact 16× 16 representation allows us to model high-resolution videos. In our framework, we learn the lip movements in the quantized space on the newly collected 4K Talking Faces (4KTF) dataset. Our approach is speaker agnostic and can handle various languages and voices. We benchmark our technique against several competitive works and show that we can achieve a remarkable 64-times more pixels than the current state-of-the-art! Our supplementary demo video depicts additional qualitative results, comparisons, and several real-world applications, like professional movie editing enabled by our model.

Bias mitigation in machine learning models is imperative, yet challenging. While several approaches have been proposed, one view towards mitigating bias is through adversarial learning. A discriminator is used to identify the bias attributes such as gender, age or race in question. This discriminator is used adversarially to ensure that it cannot distinguish the bias attributes. The main drawback in such a model is that it directly introduces a trade-off with accuracy as the features that the discriminator deems to be sensitive for discrimination of bias could be correlated with classification. In this work we solve the problem. We show that a biased discriminator can actually be used to improve this bias-accuracy tradeoff. Specifically, this is achieved by using a feature masking approach using the discriminator's gradients. We ensure that the features favoured for the bias discrimination are de-emphasized and the unbiased features are enhanced during classification. We show that this simple approach works well to reduce bias as well as improve accuracy significantly. We evaluate the proposed model on standard benchmarks. We improve the accuracy of the adversarial methods while maintaining or even improving the unbiasness and also outperform several other recent methods.

Lipreading is a difficult task, even for humans. And synthesizing the original speech waveform from lipreading makes it even a more challenging problem. Towards this end, we present a deep learning framework that can be trained end-to-end to learn the mapping between the auditory and visual signals. In particular, in this paper, our interest is to design a model that can efficiently predict the speech signal in a given silent talking-face video. The proposed framework generates a speech signal by mapping the video frames in a sequence of feature vectors. However, unlike some recent methods that adopt a sequence-to-sequence approach for translation from the frame stream to the audio stream, we posit it as an analogy learning problem between the two modalities. In which each frame is mapped to the corresponding speech segment via a deep audio-visual analogy framework. We predict plausible audio stream by training adversarially against a discriminator network. Our experiments, both qualitative and quantitative, on the publicly available GRID dataset show that the proposed method outperforms prior work on existing evaluation benchmarks. Our user studies confirm that our generated samples are more natural and closely match the ground truth speech signal.

Understanding the lip movement and inferring the speech from it is notoriously difficult for the common person. The task of accurate lip-reading gets help from various cues of the speaker and its contextual or environmental setting. Every speaker has a different accent and speaking style, which can be inferred from their visual and speech features. This work aims to understand the correlation/mapping between speech and the sequence of lip movement of individual speakers in an unconstrained and large vocabulary. We model the frame sequence as a distribution of features from the transformer in an autoencoder setting and learn the embeddings jointly that exploits temporal properties of both audio and video. We learn temporal synchronization using deep metric learning, which guides the decoder to generate speech in sync with input lip movements. The predictive posterior thus gives us the generated speech in speaker speaking style. We have trained our model on the Grid and Lip2Wav Chemistry lecture dataset to evaluate single speaker natural speech generation tasks from lip movement in an unconstrained natural setting. Extensive evaluation using various qualitative and quantitative metrics with human evaluation also shows that our method outperforms on Lip2Wav Chemistry dataset (large vocabulary in an unconstrained setting) by a good margin across almost all evaluation metrics and marginally outperforms the state-of-the-art on GRID dataset.

Shopping experience on any e-commerce website is largely driven by the content customers interact with. The large volume of diverse content on e-commerce platforms, and the advances in machine learning, pose unique opportunities for gathering insights through content understanding and applying these insights to generate content better shopper experience. The purpose of the first edition of this workshop was to bring together researchers from industry and academia on questions surrounding e-commerce content understanding and generation.

In this paper, we explore an interesting question of what can be obtained from an 8x8 pixel video sequence. Surprisingly, it turns out to be quite a lot. We show that when we process this 8x8 video with the right set of audio and image priors, we can obtain a full-length, 256x256 video. We achieve this 32x scaling of an extremely low-resolution input using our novel audio-visual upsampling network. The audio prior helps to recover the elemental facial details and precise lip shapes and a single high-resolution target identity image prior provides us with rich appearance details. Our approach is an end-to-end multi-stage framework. The first stage produces a coarse intermediate output video that can be then used to animate single target identity image and generate realistic, accurate and high-quality outputs. Our approach is simple and performs exceedingly well (an 8x improvement in FID score) compared to previous super-resolution methods. We also extend our model to talking-face video compression, and show that we obtain a 3.5x improvement in terms of bits/pixel over the previous state-of-the-art. The results from our network are thoroughly analyzed through extensive ablation experiments (in the paper and supplementary material). We also provide the demo video along with code and models on our http://cvit.iiit.ac.in/research/projects/cvit-projects/talking-face-video-upsampling.

In this work, we address the problem of generating speech from silent lip videos for any speaker in the wild. In stark contrast to previous works, our method (i) is not restricted to a fixed number of speakers, (ii) does not explicitly impose constraints on the domain or the vocabulary and (iii) deals with videos that are recorded in the wild as opposed to within laboratory settings. The task presents a host of challenges, with the key one being that many features of the desired target speech, like voice, pitch and linguistic content, cannot be entirely inferred from the silent face video. In order to handle these stochastic variations, we propose a new VAE-GAN architecture that learns to associate the lip and speech sequences amidst the variations. With the help of multiple powerful discriminators that guide the training process, our generator learns to synthesize speech sequences in any voice for the lip movements of any person. Extensive experiments on multiple datasets show that we outperform all baselines by a large margin. Further, our network can be fine-tuned on videos of specific identities to achieve a performance comparable to single-speaker models that are trained on $4\times$ more data. We conduct numerous ablation studies to analyze the effect of different modules of our architecture. We also provide a demo video that demonstrates several qualitative results along with the code and trained models on our website http://cvit.iiit.ac.in/research/projects/cvit-projects/lip-to-speech-synthesis.

Deep learning models generally learn the biases present in the training data. Researchers have proposed several approaches to mitigate such biases and make the model fair. Bias mitigation techniques assume that a sufficiently large number of training examples are present. However, we observe that if the training data is limited, then the effectiveness of bias mitigation methods is severely degraded. In this paper, we propose a novel approach to address this problem. Specifically, we adapt self-supervision and self-distillation to reduce the impact of biases on the model in this setting. Self-supervision and self-distillation are not used for bias mitigation. However, through this work, we demonstrate for the first time that these techniques are very effective in bias mitigation. We empirically show that our approach can significantly reduce the biases learned by the model. Further, we experimentally demonstrate that our approach is complementary to other bias mitigation strategies. Our approach significantly improves their performance and further reduces the model biases in the limited data regime. Specifically, on the L-CIFAR-10S skewed dataset, our approach significantly reduces the bias score of the baseline model by 78.22% and outperforms it in terms of accuracy by a significant absolute margin of 8.89%. It also significantly reduces the bias score for the state-of-the-art domain independent bias mitigation method by 59.26% and improves its performance by a significant absolute margin of 7.08%.

Deep learning models suffer from catastrophic forgetting when trained in an incremental learning setting. In this work, we propose a novel approach to address the task incremental learning problem, which involves training a model on new tasks that arrive in an incremental manner. The task incremental learning problem becomes even more challenging when the test set contains classes that are not part of the train set, i.e., a task incremental generalized zero-shot learning problem. Our approach can be used in both the zero-shot and non zero-shot task incremental learning settings. Our proposed method uses weight rectifications and affine transformations in order to adapt the model to different tasks that arrive sequentially. Specifically, we adapt the network weights to work for new tasks by "rectifying" the weights learned from the previous task. We learn these weight rectifications using very few parameters. We additionally learn affine transformations on the outputs generated by the network in order to better adapt them for the new task. We perform experiments on several datasets in both zero-shot and non zero-shot task incremental learning settings and empirically show that our approach achieves state-of-the-art results. Specifically, our approach outperforms the state-of-the-art non zero-shot task incremental learning method by over 5% on the CIFAR-100 dataset. Our approach also significantly outperforms the state-of-the-art task incremental generalized zero-shot learning method by absolute margins of 6.91% and 6.33% for the AWA1 and CUB datasets, respectively. We validate our approach using various ablation studies.

There have been a number of techniques that have demonstrated the generation of multimedia data for one modality at a time using GANs, such as the ability to generate images, videos, and audio. However, so far, the task of multi-modal generation of data, specifically for audio and videos both, has not been sufficiently well-explored. Towards this, we propose a method that demonstrates that we are able to generate naturalistic samples of video and audio data by the joint correlated generation of audio and video modalities. The proposed method uses multiple discriminators to ensure that the audio, video, and the joint output are also indistinguishable from real-world samples. We present a dataset for this task and show that we are able to generate realistic samples. This method is validated using various standard metrics such as Inception Score, Frechet Inception Distance (FID) and through human evaluation.

Understanding the relationship between the auditory and visual signals is crucial for many different applications ranging from computer-generated imagery (CGI) and video editing automation to assisting people with hearing or visual impairments. However, this is challenging since the distribution of both audio and visual modality is inherently multi-modal. Therefore, most of the existing methods ignore the multimodal aspect and assume that there only exists a deterministic one-to-one mapping between the two modalities. It can lead to low-quality predictions as the model collapses to optimizing the average behavior rather than learning the full data distributions. In this paper, we present a stochastic model for generating speech in a silent video. The proposed model combines recurrent neural networks and variational deep generative models to learn the auditory signal’s conditional distribution given the visual signal. We demonstrate the performance of our model on the GRID dataset based on standard benchmarks.

We propose a novel approach for class incremental online learning in a limited data setting. This problem setting is challenging because of the following constraints: (1) Classes are given incrementally, which necessitates a class incremental learning approach; (2) Data for each class is given in an online fashion, i.e., each training example is seen only once during training; (3) Each class has very few training examples; and (4) We do not use or assume access to any replay/memory to store data from previous classes. Therefore, in this setting, we have to handle twofold problems of catastrophic forgetting and overfitting. In our approach, we learn robust representations that are generalizable across tasks without suffering from the problems of catastrophic forgetting and overfitting to accommodate future classes with limited samples. Our proposed method leverages the meta-learning framework with knowledge consolidation. The meta-learning framework helps the model for rapid learning when samples appear in an online fashion. Simultaneously, knowledge consolidation helps to learn a robust representation against forgetting under online updates to facilitate future learning. Our approach significantly outperforms other methods on several benchmarks.

We aim to solve the highly challenging task of generating continuous sign language videos solely from speech segments for the first time. Recent efforts in this space have focused on generating such videos from human-annotated text transcripts without considering other modalities. However, replacing speech with sign language proves to be a practical solution while communicating with people suffering from hearing loss. Therefore, we eliminate the need of using text as input and design techniques that work for more natural, continuous, freely uttered speech covering an extensive vocabulary. Since the current datasets are inadequate for generating sign language directly from speech, we collect and release the first Indian sign language dataset comprising speech-level annotations, text transcripts, and the corresponding sign-language videos. Next, we propose a multi-tasking transformer network trained to generate signer's poses from speech segments. With speech-to-text as an auxiliary task and an additional cross-modal discriminator, our model learns to generate continuous sign pose sequences in an end-to-end manner. Extensive experiments and comparisons with other baselines demonstrate the effectiveness of our approach. We also conduct additional ablation studies to analyze the effect of different modules of our network. A demo video containing several results is attached to the supplementary material.

In this work, we re-think the task of speech enhancement in unconstrained real-world environments. Current state- of-the-art methods use only the audio stream and are limited in their performance in a wide range of real-world noises. Recent works using lip movements as additional cues improve the quality of generated speech over "audio-only " methods. But, these methods cannot be used for several applications where the visual stream is unreliable or completely absent. We propose a new paradigm for speech enhancement by exploiting recent breakthroughs in speech- driven lip synthesis. Using one such model as a teacher network, we train a robust student network to produce accurate lip movements that mask away the noise, thus acting as a "visual noise filter". The intelligibility of the speech enhanced by our pseudo-lip approach is comparable ( <; 3% difference) to the case of using real lips. This implies that we can exploit the advantages of using lip movements even in the absence of a real video stream. We rigorously evaluate our model using quantitative metrics as well as human evaluations. Additional ablation studies and a demo video on our website containing qualitative comparisons and results clearly illustrate the effectiveness of our approach.

One of the major limitations of deep learning models is that they face catastrophic forgetting in an incremental learning scenario. There have been several approaches proposed to tackle the problem of incremental learning. Most of these methods are based on knowledge distillation and do not adequately utilize the information provided by older task models, such as uncertainty estimation in pre-dictions. The predictive uncertainty provides the distributional information can be applied to mitigate catastrophic forgetting in a deep learning framework. In the proposed work, we consider a Bayesian formulation to obtain the data and model uncertainties. We also incorporate self-attention framework to address the incremental learning problem. We define distillation losses in terms of aleatoric uncertainty and self-attention. In the proposed work, we investigate different ablation analyses on these losses. Furthermore, we are able to obtain better results in terms of accuracy on standard benchmarks.

Unsupervised Domain adaptation methods solve the adaptation problem for an unlabeled target set, assuming that the source dataset is available with all labels. However, the availability of actual source samples is not always possible in practical cases. It could be due to memory constraints, privacy concerns, and challenges in sharing data. This practical scenario creates a bottleneck in the domain adaptation problem. This paper addresses this challenging scenario by proposing a domain adaptation technique that does not need any source data. Instead of the source data, we are only provided with a classifier that is trained on the source data. Our proposed approach is based on a generative framework, where the trained classifier is used for generating samples from the source classes. We learn the joint distribution of data by using the energy-based modeling of the trained classifier. At the same time, a new classifier is also adapted for the target domain. We perform various ablation analysis under different experimental setups and demonstrate that the proposed approach achieves better results than the baseline models in this extremely novel scenario.

In this paper, we propose an approach to improve few-shot classification performance using a composite rotation based auxiliary task. Few-shot classification methods aim to produce neural networks that perform well for classes with a large number of training samples and classes with less number of training samples. They employ techniques to enable the network to produce highly discriminative features that are also very generic. Generally, the better the quality and generic-nature of the features produced by the network, the better is the performance of the network on few-shot learning. Our approach aims to train networks to produce such features by using a self-supervised auxiliary task. Our proposed composite rotation based auxiliary task performs rotation at two levels, i.e., rotation of patches inside the image (inner rotation) and rotation of the whole image (outer rotation) and assigns one out of 16 rotation classes to the modified image. We then simultaneously train for the composite rotation prediction task along with the original classification task, which forces the network to learn high-quality generic features that help improve the few-shot classification performance. We experimentally show that our approach performs better than existing few-shot learning methods on multiple benchmark datasets.

Learning from a few examples is an important practical aspect of training classifiers. Various works have examined this aspect quite well. However, all existing approaches assume that the few examples provided are always correctly labeled. This is a strong assumption, especially if one considers the current techniques for labeling using crowd-based labeling services. We address this issue by proposing a novel robust few-shot learning approach. Our method relies on generating robust prototypes from a set of few examples. Specifically, our method refines the class prototypes by producing hybrid features from the support examples of each class. The refined prototypes help to classify the query images better. Our method can replace the evaluation phase of any few-shot learning method that uses a nearest neighbor prototype-based evaluation procedure to make them robust. We evaluate our method on standard mini-ImageNet and tiered-ImageNet datasets. We perform experiments with various label corruption rates in the support examples of the few-shot classes. We obtain significant improvement over widely used few-shot learning methods that suffer significant performance degeneration in the presence of label noise. We finally provide extensive ablation experiments to validate our method.

In this paper, we propose a novel approach for generalized zero-shot learning in a multi-modal setting, where we have novel classes ofaudio/video during testing that are not seen during training. We use the semantic relatedness oftext embeddings as a means for zero-shot learning by aligning audio and video embeddings with the corresponding class label text feature space. Our approach uses a cross-modal decoder and a composite triplet loss. The cross-modal decoder enforces a constraint that the class label text features can be reconstructed from the audio and video embeddings of data points. This helps the audio and video embeddings to move closer to the class label text embedding. The composite triplet loss makes use of the audio, video, and text embeddings. It helps bring the embeddings from the same class closer and push away the embeddings from different classes in a multi-modal setting. This helps the network to perform better on the multi-modal zero-shot learning task. Importantly, our multi-modal zero-shot learning approach works even if a modality is missing at test time. We test our approach on the generalized zero-shot classification and retrieval tasks and show that our approach outperforms other models in the presence of a single modality as well as in the presence of multiple modalities. We validate our approach by comparing it with previous approaches and using various ablations.

In recent years, the attention mechanism has become a fairly popular concept and has proven to be successful in many machine learning applications. However, deep learning models do not employ supervision for these attention mechanisms which can improve the model's performance significantly. Therefore, in this paper, we tackle this limitation and propose a novel method to improve the attention mechanism by inducing "self-supervision". We devise a technique to generate desirable attention maps for any model that utilizes an attention module. This is achieved by examining the model's output for different regions sampled from the input and obtaining the attention probability distributions that enhance the proficiency of the model. The attention distributions thus obtained are used for supervision. We rely on the fact, that attenuation of the unimportant parts, allows a model to attend to more salient regions, thus strengthening the prediction accuracy. The quantitative and qualitative results published in this paper show that this method successfully improves the attention mechanism as well as the model's accuracy. In addition to the task of Visual Question Answering(VQA), we also show results on the task of Image classification and Text classification to prove that our method can be generalized to any vision and language model that uses an attention module.

A great number of situational comedies (sitcoms) are being regularly made and the task of adding laughter tracks to these is a critical task. Providing an ability to be able to predict whether something will be humorous to the audience is also crucial. In this project, we aim to automate this task. Towards doing so, we annotate an existing sitcom (`Big Bang Theory') and use the laughter cues present to obtain a manual annotation for this show. We provide detailed analysis for the dataset design and further evaluate various state of the art baselines for solving this task. We observe that existing LSTM and BERT based networks on the text alone do not perform as well as joint text and video or only video-based networks. Moreover, it is challenging to ascertain that the words attended to while predicting laughter are indeed humorous. Our dataset and analysis provided through this paper is a valuable resource towards solving this interesting semantic and practical task. As an additional contribution, we have developed a novel model for solving this task that is a multi-modal self-attention based model that outperforms currently prevalent models for solving this task. The project page for our paper is https://delta-lab-iitk.github.io/Multimodal-Humor-Dataset/.

Hatching is a common method used by artists to accentuate the third dimension of a sketch, and to illuminate the scene. Our system SHAD3S1attempts to compete with a human at hatching generic three-dimensional (3D) shapes, and also tries to assist her in a form exploration exercise. The novelty of our approach lies in the fact that we make no assumptions about the input other than that it represents a 3D shape, and yet, given a contextual information of illumination and texture, we synthesise an accurate hatch pattern over the sketch, without access to 3D or pseudo 3D. In the process, we contribute towards a) a cheap yet effective method to synthesise a sufficiently large high fidelity dataset, pertinent to task; b) creating a pipeline with conditional generative adversarial network (CGAN); and c) creating an interactive utility with GIMP, that is a tool for artists to engage with automated hatching or a form-exploration exercise. User evaluation of the tool suggests that the model performance does generalise satisfactorily over diverse input, both in terms of style as well as shape. A simple comparison of inception scores suggest that the generated distribution is as diverse as the ground truth.

Understanding and explaining deep learning models is an imperative task. Towards this, we propose a method that obtains gradient-based certainty estimates that also provide visual attention maps. Particularly, we solve for visual question answering task. We incorporate modern probabilistic deep learning methods that we further improve by using the gradients for these estimates. These have two-fold benefits: a) improvement in obtaining the certainty estimates that correlate better with misclassified samples and b) improved attention maps that provide state-of-the-art results in terms of correlation with human attention regions. The improved attention maps result in consistent improvement for various methods for visual question answering. Therefore, the proposed technique can be thought of as a tool for obtaining improved certainty estimates and explanations for deep learning models. We provide detailed empirical analysis for the visual question answering task on all standard benchmarks and comparison with state of the art methods.

In this paper, we aim to obtain improved attention for a visual question answering (VQA) task. It is challenging to provide supervision for attention. An observation we make is that visual explanations as obtained through class activation mappings (specifically Grad-CAM) that are meant to explain the performance of various networks could form a means of supervision. However, as the distributions of attention maps and that of Grad-CAMs differ, it would not be suitable to directly use these as a form of supervision. Rather, we propose the use of a discriminator that aims to distinguish samples of visual explanation and attention maps. The use of adversarial training of the attention regions as a two-player game between attention and explanation serves to bring the distributions of attention maps and visual explanations closer. Significantly, we observe that providing such a means of supervision also results in attention maps that are more closely related to human attention resulting in a substantial improvement over baseline stacked attention network (SAN) models. It also results in a good improvement in rank correlation metric on the VQA task. This method can also be combined with recent MCB based methods and results in consistent improvement. We also provide comparisons with other means for learning distributions such as based on Correlation Alignment (Coral), Maximum Mean Discrepancy (MMD) and Mean Square Error (MSE) losses and observe that the adversarial loss outperforms the other forms of learning the attention maps. Visualization of the results also confirms our hypothesis that attention maps improve using this form of supervision.

Humans involuntarily tend to infer parts of the conversation from lip movements when the speech is absent or corrupted by external noise. In this work, we explore the task of lip to speech synthesis, i.e., learning to generate natural speech given only the lip movements of a speaker. Acknowledging the importance of contextual and speaker-specific cues for accurate lip-reading, we take a different path from existing works. We focus on learning accurate lip sequences to speech mappings for individual speakers in unconstrained, large vocabulary settings. To this end, we collect and release a large-scale benchmark dataset, the first of its kind, specifically to train and evaluate the single-speaker lip to speech task in natural settings. We propose a novel approach with key design choices to achieve accurate, natural lip to speech synthesis in such unconstrained scenarios for the first time. Extensive evaluation using quantitative, qualitative metrics and human evaluation shows that our method is four times more intelligible than previous works in this space.

Convolutional layers are a major driving force behind the successes of deep learning. Pointwise convolution (PWC) is a 1 × 1 convolutional filter that is primarily used for parameter reduction. However, the PWC ignores the spatial information around the points it is processing. This design is by choice, in order to reduce the overall parameters and computations. However, we hypothesize that this shortcoming of PWC has a significant impact on the network performance. We propose an alternative design for pointwise convolution, which uses spatial information from the input efficiently. Our design significantly improves the performance of the networks without substantially increasing the number of parameters and computations. We experimentally show that our design results in significant improvement in the performance of the network for classification as well as detection.

This work presents a novel training technique for deep neural networks that makes use of additional data from a distribution that is different from that of the original input data. This technique aims to reduce overfitting and improve the generalization performance of the network. Our proposed technique, namely Passive Batch Injection Training Technique (PBITT), even reduces the level of overfitting in networks that already use the standard techniques for reducing overfitting such as L2regularization and batch normalization, resulting in significant accuracy improvements. Passive Batch Injection Training Technique (PBITT) introduces a few passive mini-batches into the training process that contain data from a distribution that is different from the input data distribution. This technique does not increase the number of parameters in the final model and also does not increase the inference (test) time but still improves the performance of deep CNNs. To the best of our knowledge, this is the first work that makes use of different data distribution to aid the training of convolutional neural networks (CNNs). We thoroughly evaluate the proposed approach on standard architectures: VGG, ResNet, and WideResNet, and on several popular datasets: CIFAR-10, CIFAR-100, SVHN, and ImageNet. We observe consistent accuracy improvement by using the proposed technique. We also show experimentally that the model trained by our technique generalizes well to other tasks such as object detection on the MS-COCO dataset using Faster R-CNN. We present extensive ablations to validate the proposed approach. Our approach improves the accuracy of VGG-16 by a significant margin of 2.1% over the CIFAR-100 dataset.

Convolution operation in deep convolutional neural networks is the most computationally expensive as compared to other operations. Most of the model computation (FLOPS) in the deep architecture belong to convolution operation. In this paper, we are proposing a novel skip convolution operation that employs significantly fewer computation as compared to the traditional one without sacrificing model accuracy. Skip convolution operation produces structured sparsity in the output feature maps without requiring sparsity in the model parameters for computation reduction. The existing convolution operation performs the redundant computation for object feature representation while the proposed convolution skips redundant computation. Our empirical evaluation for various deep models (VGG, ResNet, MobileNet, and Faster R-CNN) over various benchmarked datasets (CIFAR-10, CIFAR-100, ImageNet, and MS-COCO) show that skip convolution reduces the computation significantly while preserving feature representational capacity. The proposed approach is model-agnostic and can be applied over any architecture. The proposed approach does not require a pretrained model and does train from scratch. Hence we achieve significant computation reduction at training and test time. We are also able to reduce computation in an already compact model such as MobileNet using skip convolution. We also show empirically that the proposed convolution works well for other tasks such as object detection. Therefore, SkipConv can be a widely usable and efficient way of reducing computation in deep CNN models.

The ability to envisage the visual of a talking face based just on hearing a voice is a unique human capability. There have been a number of works that have solved for this ability recently. We differ from these approaches by enabling a variety of talking face generations based on single audio input. Indeed, just having the ability to generate a single talking face would make a system almost robotic in nature. In contrast, our unsupervised stochastic audio-to-video generation model allows for diverse generations from a single audio input. Particularly, we present an unsupervised stochastic audio-to-video generation model that can capture multiple modes of the video distribution. We ensure that all the diverse generations are plausible. We do so through a principled multi-modal variational autoencoder framework. We demonstrate its efficacy on the challenging LRW and GRID datasets and demonstrate performance better than the baseline, while having the ability to generate multiple diverse lip synchronized videos.

Convolutional Neural Networks (CNNs) have been successfully applied for relative camera pose estimation from labeled image-pair data, without requiring any handengineered features, camera intrinsic parameters or depth information. The trained CNN can be utilized for performing pose based visual servo control (PBVS). One of the ways to improve the quality of visual servo output is to improve the accuracy of the CNN for estimating the relative pose estimation. With a given state-of-the-art CNN for relative pose regression, how can we achieve an improved performance for visual servo control? In this paper, we explore switching of CNNs to improve the precision of visual servo control. The idea of switching a CNN is due to the fact that the dataset for training a relative camera pose regressor for visual servo control must contain variations in relative pose ranging from a very small scale to eventually a larger scale. We found that, training two different instances of the CNN, one for large-scale-displacements (LSD) and another for small-scale-displacements (SSD) and switching them during the visual servo execution yields better results than training a single CNN with the combined LSD+SSD data. However, it causes extra storage overhead and switching decision is taken by a manually set threshold which may not be optimal for all the scenes. To eliminate these drawbacks, we propose an efficient switching strategy based on model agnostic meta learning (MAML) algorithm. In this, a single model is trained to learn parameters which are simultaneously good for multiple tasks, namely a binary classification for switching decision, a 6DOF pose regression for LSD data and also a 6DOF pose regression for SSD data. The proposed approach performs far better than the naive approach, while storage and run-time overheads are almost negligible.

We present sentence aligned parallel corpora across 10 Indian Languages - Hindi, Telugu, Tamil, Malayalam, Gujarati, Urdu, Bengali, Oriya, Marathi, Punjabi, and English - many of which are categorized as low resource. The corpora are compiled from online sources which have content shared across languages. The corpora presented significantly extends present resources that are either not large enough or are restricted to a specific domain (such as health). We also provide a separate test corpus compiled from an independent online source that can be independently used for validating the performance in 10 Indian languages. Alongside, we report on the methods of constructing such corpora using tools enabled by recent advances in machine translation and cross-lingual retrieval using deep neural network based methods.

We present the problem of Visually Precise Query (VPQ) generation which enables a more intuitive match between a user's information need and an e-commerce site's product description. Given an image of a fashion item, what is the most optimum search query that will retrieve the exact same or closely related product(s) with high probability. In this paper we introduce the task of VPQ generation which takes a product image and its title as its input and provides aword level extractive summary of the title, containing a list of salient attributes, which can now be used as a query to search for similar products. We collect a large dataset of fashion images and their titles and merge it with an existing research dataset which was created for a different task. Given the image and title pair, VPQ problem is posed as identifying a non-contiguous collection of spans within the title. We provide a dataset of around 400K image, title and corresponding VPQ entries and release it to the research community. We provide a detailed description of the data collection process as well as discuss the future direction of research for the problem introduced in this work. We provide the standard text as well as visual domain baseline comparisons and also provide multi-modal baseline models to analyze the task introduced in this work. Finally, we propose a hybrid fusion model which promises to be the direction of research in the multi-modal community.

In this work, we investigate the problem of lip-syncing a talking face video of an arbitrary identity to match a target speech segment. Current works excel at producing accurate lip movements on a static image or videos of specific people seen during the training phase. However, they fail to accurately morph the lip movements of arbitrary identities in dynamic, unconstrained talking face videos, resulting in significant parts of the video being out-of-sync with the new audio. We identify key reasons pertaining to this and hence resolve them by learning from a powerful lip-sync discriminator. Next, we propose new, rigorous evaluation benchmarks and metrics to accurately measure lip synchronization in unconstrained videos. Extensive quantitative evaluations on our challenging benchmarks show that the lip-sync accuracy of the videos generated by our Wav2Lip model is almost as good as real synced videos. We provide a demo video clearly showing the substantial impact of our Wav2Lip model, and also publicly release the code, models, and evaluation benchmarks on our website.

Training Neural Ordinary Differential Equations (ODEs) is often computationally expensive. Indeed, computing the forward pass of such models involves solving an ODE which can become arbitrarily complex during training. Recent works have shown that regularizing the dynamics of the ODE can partially alleviate this. In this paper we propose a new regularization technique: randomly sampling the end time of the ODE during training. The proposed regularization is simple to implement, has negligible overhead and is effective across a wide variety of tasks. Further, the technique is orthogonal to several other methods proposed to regularize the dynamics of ODEs and as such can be used in conjunction with them. We show through experiments on normalizing flows, time series models and image recognition that the proposed regularization can significantly decrease training time and even improve performance over baseline models.

In this work, we propose a modeling technique for jointly training image and video generation models by simultaneously learning to map latent variables with a fixed prior onto real images and interpolate over images to generate videos. The proposed approach models the variations in representations using residual vectors encoding the change at each time step over a summary vector for the entire video. We utilize the technique to jointly train an image generation model with a fixed prior along with a video generation model lacking constraints such as disentanglement. The joint training enables the image generator to exploit temporal information while the video generation model learns to flexibly share information across frames. Moreover, experimental results verify our approach's compatibility with pre-training on videos or images and training on datasets containing a mixture of both. A comprehensive set of quantitative and qualitative evaluations reveal the improvements in sample quality and diversity over both video generation and image generation baselines. We further demonstrate the technique's capabilities of exploiting similarity in features across frames by applying it to a model based on decomposing the video into motion and content. The proposed model allows minor variations in content across frames while maintaining the temporal dependence through latent vectors encoding the pose or motion features.

Generating natural questions from an image is a semantic task that requires using vision and language modalities to learn multimodal representations. Images can have multiple visual and language cues such as places, captions, and tags. In this paper, we propose a principled deep Bayesian learning framework that combines these cues to produce natural questions. We observe that with the addition of more cues and by minimizing uncertainty in the among cues, the Bayesian network becomes more confident. We propose a Minimizing Uncertainty of Mixture of Cues (MUMC), that minimizes uncertainty present in a mixture of cues experts for generating probabilistic questions. This is a Bayesian framework and the results show a remarkable similarity to natural questions as validated by a human study. We observe that with the addition of more cues and by minimizing uncertainty among the cues, the Bayesian framework becomes more confident. Ablation studies of our model indicate that a subset of cues is inferior at this task and hence the principled fusion of cues is preferred. Further, we observe that the proposed approach substantially improves over state-of-the-art benchmarks on the quantitative metrics (BLEU-n, METEOR, ROUGE, and CIDEr). Here we provide project link for Deep Bayesian VQG https: //delta-lab-iitk.github.io/BVQG/.

In this paper, we propose a method to obtain robust explanations for visual question answering(VQA) that correlate well with the answers. Our model explains the answers obtained through a VQA model by providing visual and textual explanations. The main challenges that we address are i) Answers and textual explanations obtained by current methods are not well correlated and ii) Current methods for visual explanation do not focus on the right location for explaining the answer. We address both these challenges by using a collaborative correlated module which ensures that even if we do not train for noise based attacks, the enhanced correlation ensures that the right explanation and answer can be generated. We further show that this also aids in improving the generated visual and textual explanations. The use of the correlated module can be thought of as a robust method to verify if the answer and explanations are coherent. We evaluate this model using VQA-X dataset. We observe that the proposed method yields better textual and visual justification that supports the decision. We showcase the robustness of the model against a noise-based perturbation attack using corresponding visual and textual explanations. A detailed empirical analysis is shown.

Convolution Neural Networks (CNN) have been extremely successful in solving intensive computer vision tasks. The convolutional filters used in CNNs have played a major role in this success, by extracting useful features from the inputs. Recently researchers have tried to boost the performance of CNNs by re-calibrating the feature maps produced by these filters, e.g., Squeeze-and-Excitation Networks (SENets). These approaches have achieved better performance by Exciting up the important channels or feature maps while diminishing the rest. However, in the process, architectural complexity has increased. We propose an architectural block that introduces much lower complexity than the existing methods of CNN performance boosting while performing significantly better than them. We carry out experiments on the CIFAR, ImageNet and MS-COCO datasets, and show that the proposed block can challenge the state-of-the-art results. Our method boosts the ResNet-50 architecture to perform comparably to the ResNet-152 architecture, which is a three times deeper network, on classification. We also show experimentally that our method is not limited to classification but also generalizes well to other tasks such as object detection.

Researchers have proposed various activation functions. These activation functions help the deep network to learn non-linear behavior with a significant effect on training dynamics and task performance. The performance of these activations also depends on the initial state of the weight parameters, i.e., different initial state leads to a difference in the performance of a network. In this paper, we have proposed a cooperative initialization for training the deep network using ReLU activation function to improve the network performance. Our approach uses multiple activation functions in the initial few epochs for the update of all sets of weight parameters while training the network. These activation functions cooperate to overcome their drawbacks in the update of weight parameters, which in effect learn better "feature representation" and boost the network performance later. Cooperative initialization based training also helps in reducing the overfitting problem and does not increase the number of parameters, inference (test) time in the final model while improving the performance. Experiments show that our approach outperforms various baselines and, at the same time, performs well over various tasks such as classification and detection. The Top-1 classification accuracy of the model trained using our approach improves by 2.8% for VGG-16 and 2.1% for ResNet-56 on CIFAR-100 dataset.

We present a filter correlation based model compression approach for deep convolutional neural networks. Our approach iteratively identifies pairs of filters with the largest pairwise correlations and drops one of the filters from each such pair. However, instead of discarding one of the filters from each such pair naïvely, the model is re-optimized to make the filters in these pairs maximally correlated, so that discarding one of the filters from the pair results in minimal information loss. Moreover, after discarding the filters in each round, we further finetune the model to recover from the potential small loss incurred by the compression. We evaluate our proposed approach using a comprehensive set of experiments and ablation studies. Our compression method yields state-of-the-art FLOPs compression rates on various benchmarks, such as LeNet-5, VGG-16, and ResNet-50,56, while still achieving excellent predictive performance for tasks such as object detection on benchmark datasets.

Line art is arguably one of the fundamental and versatile modes of expression. We propose a pipeline for a robot to look at a grayscale line art and redraw it. The key novel elements of our pipeline are: a) we propose a novel task of mimicking line drawings, b) to solve the pipeline we modify the Quick-draw dataset and obtain supervised training for converting a line drawing into a series of strokes c) we propose a multi-stage segmentation and graph interpretation pipeline for solving the problem. The resultant method has also been deployed on a CNC plotter as well as a robotic arm. We have trained several variations of the proposed methods and evaluate these on a dataset obtained from Quick-draw. Through the best methods we observe an accuracy of around 98% for this task, which is a significant improvement over the baseline architecture we adapted from. This therefore allows for deployment of the method on robots for replicating line art in a reliable manner. We also show that while the rule-based vectorization methods do suffice for simple drawings, it fails for more complicated sketches, unlike our method which generalizes well to more complicated distributions.

We present a filter pruning approach for deep model compression, using a multitask network. Our approach is based on learning a a pruner network to prune a pre-trained target network. The pruner is essentially a multitask deep neural network with binary outputs that help identify the filters from each layer of the original network that do not have any significant contribution to the model and can therefore be pruned. The pruner network has the same architecture as the original network except that it has a multitask/multi-output last layer containing binary-valued outputs (one per filter), which indicate which filters have to be pruned. The pruner's goal is to minimize the number of filters from the original network by assigning zero weights to the corresponding output feature-maps. In contrast to most of the existing methods, instead of relying on iterative pruning, our approach can prune the network (original network) in one go and, moreover, does not require specifying the degree of pruning for each layer (and can learn it instead). The compressed model produced by our approach is generic and does not need any special hardware/software support. Moreover, augmenting with other methods such as knowledge distillation, quantization, and connection pruning can increase the degree of compression for the proposed approach. We show the efficacy of our proposed approach for classification and object detection tasks.

Generative models have recently shown the ability to realistically generate data and model the distribution accurately. However, joint modeling of an image with the attribute that it is labeled with requires learning a cross modal correspondence between image and attribute data. Though the information present in a set of images and its attributes possesses completely different statistical properties altogether, there exists an inherent correspondence that is challenging to capture. Various models have aimed at capturing this correspondence either through joint modeling of a variational autoencoder or through separate encoder networks that are then concatenated. We present an alternative by proposing a bridged variational autoencoder that allows for learning cross-modal correspondence by incorporating cross-modal hallucination losses in the latent space. In comparison to the existing methods, we have found that by using a bridge connection in latent space we not only obtain better generation results, but also obtain highly parameter-efficient model which provide 40% reduction in training parameters for bimodal dataset and nearly 70% reduction for trimodal dataset. We validate the proposed method through comparison with state of the art methods and benchmarking on standard datasets.

In this paper, we aim to solve for unsupervised domain adaptation of classifiers where we have access to label information for the source domain while these are not available for a target domain. While various methods have been proposed for solving these including adversarial discriminator based methods, most approaches have focused on the entire image based domain adaptation. In an image, there would be regions that can be adapted better, for instance, the foreground object may be similar in nature. To obtain such regions, we propose methods that consider the probabilistic certainty estimate of various regions and specific focus on these during classification for adaptation. We observe that just by incorporating the probabilistic certainty of the discriminator while training the classifier, we are able to obtain state of the art results on various datasets as compared against all the recent methods. We provide a thorough empirical analysis of the method by providing ablation analysis, statistical significance test, and visualization of the attention maps and t-SNE embeddings. These evaluations convincingly demonstrate the effectiveness of the proposed approach.

We present a novel deep learning architecture in which the convolution operation leverages heterogeneous kernels. The proposed HetConv (Heterogeneous Kernel-Based Convolution) reduces the computation (FLOPs) and the number of parameters as compared to standard convolution operation while still maintaining representational efficiency. To show the effectiveness of our proposed convolution, we present extensive experimental results on the standard convolutional neural network (CNN) architectures such as VGG and ResNet. We find that after replacing the standard convolutional filters in these architectures with our proposed HetConv filters, we achieve 3X to 8X FLOPs based improvement in speed while still maintaining (and sometimes improving) the accuracy. We also compare our proposed convolutions with group/depth wise convolutions and show that it achieves more FLOPs reduction with significantly higher accuracy.

Understanding videos of people speaking across international borders is hard as audiences from different demographies do not understand the language. Such speech videos are often supplemented with language subtitles. However, these hamper the viewing experience as the attention is shared. Simple audio dubbing in a different language makes the video appear unnatural due to unsynchronized lip motion. In this paper, we propose a system for automated cross-language lip synchronization for re-dubbed videos. Our model generates superior photorealistic lip-synchronization over original video in comparison to the current re-dubbing method. With the help of a user-based study, we verify that our method is preferred over unsynchronized videos.

Understanding and explaining deep learning models is an imperative task. Towards this, we propose a method that obtains gradient-based certainty estimates that also provide visual attention maps. Particularly, we solve for visual question answering task. We incorporate modern probabilistic deep learning methods that we further improve by using the gradients for these estimates. These have two-fold benefits: a) improvement in obtaining the certainty estimates that correlate better with misclassified samples and b) improved attention maps that provide state-of-the-art results in terms of correlation with human attention regions. The improved attention maps result in consistent improvement for various methods for visual question answering. Therefore, the proposed technique can be thought of as a recipe for obtaining improved certainty estimates and explanation for deep learning models. We provide detailed empirical analysis for the visual question answering task on all standard benchmarks and comparison with state of the art methods.

While convolutional neural networks (CNN) have achieved impressive performance on various classification/recognition tasks, they typically consist of a massive number of parameters. This results in significant memory requirement as well as computational overheads. Consequently, there is a growing need for filter-level pruning approaches for compressing CNN based models that not only reduce the total number of parameters but reduce the overall computation as well. We present a new min-max framework for filter-level pruning of CNNs. Our framework, called Play and Prune (PP), jointly prunes and fine-tunes CNN model parameters, with an adaptive pruning rate, while maintaining the model's predictive performance. Our framework consists of two modules: (1) An adaptive filter pruning (AFP) module, which minimizes the number of filters in the model; and (2) A pruning rate controller (PRC) module, which maximizes the accuracy during pruning. Moreover, unlike most previous approaches, our approach allows directly specifying the desired error tolerance instead of pruning level. Our compressed models can be deployed at run-time, without requiring any special libraries or hardware. Our approach reduces the number of parameters of VGG-16 by an impressive factor of 17.5X, and number of FLOPS by 6.43X, with no loss of accuracy, significantly outperforming other state-of-the-art filter pruning methods.

Abnormal activity recognition requires detection of occurrence of anomalous events that suffer from a severe imbalance in data. In a video, normal is used to describe activities that conform to usual events while the irregular events which do not conform to the normal are referred to as abnormal. It is far more common to observe normal data than to obtain abnormal data in visual surveillance. In this paper, we propose an approach where we can obtain abnormal data by transforming normal data. This is a challenging task that is solved through a multi-stage pipeline approach. We utilize a number of techniques from unsupervised segmentation in order to synthesize new samples of data that are transformed from an existing set of normal examples. Further, this synthesis approach has useful applications as a data augmentation technique. An incrementally trained Bayesian convolutional neural network (CNN) is used to carefully select the set of abnormal samples that can be added. Finally through this synthesis approach we obtain a comparable set of abnormal samples that can be used for training the CNN for the classification of normal vs abnormal samples. We show that this method generalizes to multiple settings by evaluating it on two real world datasets and achieves improved performance over other probabilistic techniques that have been used in the past for this task.

In this paper, we solve the problem of adapting classifiers across domains. We consider the problem of domain adaptation for multi-class classification where we are provided a labeled set of examples in a source dataset and we are provided a target dataset with no supervision. In this setting, we propose an adversarial discriminator based approach. While the approach based on adversarial discriminator has been previously proposed; in this paper, we present an informed adversarial discriminator. Our observation relies on the analysis that shows that if the discriminator has access to all the information available including the class structure present in the source dataset, then it can guide the transformation of features of the target set of classes to a more structure adapted space. Using this formulation, we obtain state-of-the-art results for the standard evaluation on benchmark datasets. We further provide detailed analysis which shows that using all the labeled information results in an improved domain adaptation.

In light of the recent breakthroughs in automatic machine translation systems, we propose a novel approach that we term as "Face-to-Face Translation". As today's digital communication becomes increasingly visual, we argue that there is a need for systems that can automatically translate a video of a person speaking in language A into a target language B with realistic lip synchronization. In this work, we create an automatic pipeline for this problem and demonstrate its impact in multiple real-world applications. First, we build a working speech-to-speech translation system by bringing together multiple existing modules from speech and language. We then move towards "Face-to-Face Translation" by incorporating a novel visual module, LipGAN for generating realistic talking faces from the translated audio. Quantitative evaluation of LipGAN on the standard LRW test set shows that it significantly outperforms existing approaches across all standard metrics. We also subject our Face-to-Face Translation pipeline, to multiple human evaluations and show that it can significantly improve the overall user experience for consuming and interacting with multimodal content across languages. Code, models and demo video are made publicly available.

Convolutional neural networks (CNN) have achieved impressive performance on the wide variety of tasks (classification, detection, etc.) across multiple domains at the cost of high computational and memory requirements. Thus, leveraging CNNs for real-time applications necessitates model compression approaches that not only reduce the total number of parameters but reduce the overall computation as well. In this work, we present a stability-based approach for filter-level pruning of CNNs. We evaluate our proposed approach on different architectures (LeNet, VGG-16, ResNet, and Faster RCNN) and datasets and demonstrate its generalizability through extensive experiments. Moreover, our compressed models can be used at run-time without requiring any special libraries or hardware. Our model compression method reduces the number of FLOPS by an impressive factor of 6.03X and GPU memory footprint by more than 17X, significantly outperforming other state-of-the-art filter pruning methods.

We propose a framework for compressing state-of-the-art Single Shot MultiBox Detector (SSD). The framework addresses compression in the following stages: Sparsity Induction, Filter Selection, and Filter Pruning. In the Sparsity Induction stage, the object detector model is sparsified via an improved global threshold. In Filter Selection & Pruning stage, we select and remove filters using sparsity statistics of filter weights in two consecutive convolutional layers. This results in the model with the size smaller than most existing compact architectures. We evaluate the performance of our framework with multiple datasets and compare over multiple methods. Experimental results show that our method achieves state-of-the-art compression of 6.7X and 4.9X on PASCAL VOC dataset on models SSD300 and SSD512 respectively. We further show that the method produces maximum compression of 26X with SSD512 on German Traffic Sign Detection Benchmark (GTSDB). Additionally, we also empirically show our method's adaptability for classification based architecture VGG16 on datasets CIFAR and German Traffic Sign Recognition Benchmark (GTSRB) achieving a compression rate of 125X and 200X with the reduction in flops by 90.50% and 96.6% respectively with no loss of accuracy. In addition to this, our method does not require any special libraries or hardware support for the resulting compressed models.

Generative adversarial networks (GANs) while being very versatile in realistic image synthesis, still are sensitive to the input distribution. Given a set of data that has an imbalance in the distribution, the networks are susceptible to missing modes and not capturing the data distribution. While various methods have been tried to improve training of GANs, these have not addressed the challenges of covering the full data distribution. Specifically, a generator is not penalized for missing a mode. We show that these are therefore still susceptible to not capturing the full data distribution. In this paper, we propose a simple approach that combines an encoder based objective with novel loss functions for generator and discriminator that improves the solution in terms of capturing missing modes. We validate that the proposed method results in substantial improvements through its detailed analysis on toy and real datasets. The quantitative and qualitative results demonstrate that the proposed method improves the solution for the problem of missing modes and improves training of GANs.

We propose MAD-GAN, an intuitive generalization to the Generative Adversarial Networks (GANs) and its conditional variants to address the well known problem of mode collapse. First, MAD-GAN is a multi-agent GAN architecture incorporating multiple generators and one discriminator. Second, to enforce that different generators capture diverse high probability modes, the discriminator of MAD-GAN is designed such that along with finding the real and fake samples, it is also required to identify the generator that generated the given fake sample. Intuitively, to succeed in this task, the discriminator must learn to push different generators towards different identifiable modes. We perform extensive experiments on synthetic and real datasets and compare MAD-GAN with different variants of GAN. We show high quality diverse sample generations for challenging tasks such as image-to-image translation and face generation. In addition, we also show that MAD-GAN is able to disentangle different modalities when trained using highly challenging diverse-class dataset (e.g. dataset with images of forests, icebergs, and bedrooms). In the end, we show its efficacy on the unsupervised feature representation task.

In this paper we aim to answer questions based on images when provided with a dataset of question-answer pairs for a number of images during training. A number of methods have focused on solving this problem by using image based attention. This is done by focusing on a specific part of the image while answering the question. Humans also do so when solving this problem. However, the regions that the previous systems focus on are not correlated with the regions that humans focus on. The accuracy is limited due to this drawback. In this paper, we propose to solve this problem by using an exemplar based method. We obtain one or more supporting and opposing exemplars to obtain a differential attention region. This differential attention is closer to human attention than other image based attention methods. It also helps in obtaining improved accuracy when answering questions. The method is evaluated on challenging benchmark datasets. We perform better than other image based attention methods and are competitive with other state of the art methods that focus on both image and questions.

Generating natural questions from an image is a semantic task that requires using visual and language modality to learn multimodal representations. Images can have multiple visual and language contexts that are relevant for generating questions namely places, captions, and tags. In this paper, we propose the use of exemplars for obtaining the relevant context. We obtain this by using a Multimodal Differential Network to produce natural and engaging questions. The generated questions show a remarkable similarity to the natural questions as validated by a human study. Further, we observe that the proposed approach substantially improves over state-of-the-art benchmarks on the quantitative metrics (BLEU, METEOR, ROUGE, and CIDEr).

Our goal is to spot words in silent speech videos without explicitly recognizing the spoken words, where the lip motion of the speaker is clearly visible and audio is absent. Existing work in this domain has mainly focused on recognizing a fixed set of words in word-segmented lip videos, which limits the applicability of the learned model due to limited vocabulary and high dependency on the model's recognition performance. Our contribution is two-fold: 1) we develop a pipeline for recognition-free retrieval, and show its performance against recognition-based retrieval on a large-scale dataset and another set of out-of-vocabulary words. 2) We introduce a query expansion technique using pseudo-relevant feedback and propose a novel re-ranking method based on maximizing the correlation between spatio-temporal landmarks of the query and the top retrieval candidates. Our word spotting method achieves 35% higher mean average precision over recognition-based method on large-scale LRWdataset. Finally, we demonstrate the application of the method by word spotting in a popular speech video ("The great dictator" by Charlie Chaplin) where we show that the word retrieval can be used to understand what was spoken perhaps in the silent movies.

Understanding object motions and transformations is a core problem in computer science. Modeling sequences of evolving images may provide better representations and models of motion and may ultimately be used for forecasting or simulation. Diagrammatic Abstract Reasoning is an avenue in which diagrams evolve in complex patterns and one needs to infer the underlying pattern sequence and generate the next image in the sequence. For this, we develop a novel Contextual Generative Adversarial Network based on Recurrent Neural Networks (Context-RNN-GANs), where both the generator and the discriminator modules are based on contextual history and the adversarial discriminator guides the generator to produce realistic images for the particular time step in the image sequence. We employ the Context-RNN-GAN model (and its variants) on a novel dataset of Diagrammatic Abstract Reasoning as well as perform initial evaluations on a next-frame prediction task of videos. Empirically, we show that our Context-RNN-GAN model performs competitively with 10th-grade human performance but there is still scope for interesting improvements as compared to college-grade human performance.

Abstract A hallmark of early stage design is a number of quick‐and‐dirty sketches capturing design inspirations, model variations and alternate viewpoints of a visual concept. We present SketchSoup, a workflow that allows designers to explore the design space induced by such sketches. We take an unstructured collection of drawings as input, along with a small number of user‐provided correspondences as input. We register them using a multi‐image matching algorithm, and present them as a 2D interpolation space. By morphing sketches in this space, our approach produces plausible visualizations of shape and viewpoint variations despite the presence of sketch distortions that would prevent standard camera calibration and 3D reconstruction. In addition, our interpolated sketches can serve as inspiration for further drawings, which feed back into the design space as additional image inputs. SketchSoup thus fills a significant gap in the early ideation stage of conceptual design by allowing designers to make better informed choices before proceeding to more expensive 3D modelling and prototyping. From a technical standpoint, we describe an end‐to‐end system that judiciously combines and adapts various image processing techniques to the drawing domain—where the images are dominated not by colour, shading and texture, but by sketchy stroke contours.

In this paper, we propose subspace alignment based domain adaptation of the state of the art RCNN based object detector. The aim is to be able to achieve high quality object detection in novel, real world target scenarios without requiring labels from the target domain. While, unsupervised domain adaptation has been studied in the case of object classification, for object detection it has been relatively unexplored. In subspace based domain adaptation for objects, we need access to source and target subspaces for the bounding box features. The absence of supervision (labels and bounding boxes are absent) makes the task challenging. In this paper, we show that we can still adapt sub- spaces that are localized to the object by obtaining detections from the RCNN detector trained on source and applied on target. Then we form localized subspaces from the detections and show that subspace alignment based adaptation between these subspaces yields improved object detection. This evaluation is done by considering challenging real world datasets of PASCAL VOC as source and validation set of Microsoft COCO dataset as target for various categories.

In classification of objects substantial work has gone into improving the low level representation of an image by considering various aspects such as different features, a number of feature pooling and coding techniques and considering different kernels. Unlike these works, in this paper, we propose to enhance the semantic representation of an image. We aim to learn the most important visual components of an image and how they interact in order to classify the objects correctly. To achieve our objective, we propose a new latent SVM model for category level object classification. Starting from image-level annotations, we jointly learn the object class and its context in terms of spatial location (where) and appearance (what). Furthermore, to regularize the complexity of the model we learn the spatial and co-occurrence relations between adjacent regions, such that unlikely configurations are penalized. Experimental results demonstrate that the proposed method can consistently enhance results on the challenging Pascal VOC dataset in terms of classification and weakly supervised detection. We also show how semantic representation can be exploited for finding similar content.

In this paper we propose the use of nonuniformly resized image patch exemplars for solving low level vision problems like denoising and super-resolution. While patch-based methods have been shown to be successful for several such applications, these methods have so far assumed uniform sizes for image patches. In this paper we address this restriction. We use an integral image representation for efficient computation of the matching cost for variable-sized patches. We show that nonuniform image patch exemplars are useful in improving classic techniques for nonlocal means-based denoising and example-based superresolution. We provide refinement cues to further improve the patch size estimation. This method can be adopted for a large number of related methods and applications due to its simplicity and generality.

In this paper we propose a generic framework to incorporate unobserved auxiliary information for classifying objects and actions. This framework allows us to explicitly account for localisation and alignment of representations for generic object and action classes as latent variables. We approach this problem in the discriminative setting as learning a max-margin classifier that infers the class label along with the latent variables. Through this paper we make the following contributions a) We provide a method for incorporating latent variables into object and action classification b) We specifically account for the presence of an explicit class related subregion which can include foreground and/or background. c) We explore a way to learn a better classifier by iterative expansion of the latent parameter space. We demonstrate the performance of our approach by rigorous experimental evaluation on a number of standard object and action recognition datasets. © 2011. The copyright of this document resides with its authors.

Currently two evaluation methods of super-resolution (SR) techniques prevail: The objective Peak Signal to Noise Ratio (PSNR) and a qualitative measure based on manual visual inspection. Both of these methods are sub-optimal: The latter does not scale well to large numbers of images, while the former does not necessarily reflect the perceived visual quality. We address these issues in this paper and propose an evaluation method based on image classification. We show that perceptual image quality measures like structural similarity are not suitable for evaluation of SR methods. On the other hand a systematic evaluation using large datasets of thousands of real-world images provides a consistent comparison of SR algorithms that corresponds to perceived visual quality. We verify the success of our approach by presenting an evaluation of three recent super-resolution algorithms on standard image classification datasets.

We address the problem of recognizing actions in reallife videos. Space-time interest point-based approaches have been widely prevalent towards solving this problem. In contrast, more spatially extended features such as regions have not been so popular. The reason is, any local region based approach requires the motion flow information for a specific region to be collated temporally. This is challenging as the local regions are deformable and not well delineated from the surroundings. In this paper we address this issue by using robust tracking of regions and we show that it is possible to obtain region descriptors for classification of actions. This paper lays the groundwork for further investigation into region based approaches. Through this paper we make the following contributions a) We advocate identification of salient regions based on motion segmentation b) We adopt a state-of-the art tracker for robust tracking of the identified regions rather than using isolated space-time blocks c) We propose optical flow based region descriptors to encode the extracted trajectories in piece-wise blocks. We demonstrate the performance of our system on real-world data sets.

In this paper we investigate the challenging problem of recovering the depth layers in a scene from a single defocused observation. The problem is definitely solvable if there are multiple observations. In this paper we show that one can perceive the depth in the scene even from a single observation. We use the inhomogeneous reverse heat equation to obtain an estimate of the blur, thereby preserving the depth information characterized by the defocus. However, the reverse heat equation, due to its parabolic nature, is divergent. We stabilize the reverse heat equation by considering the gradient degeneration as an effective stopping criterion. The amount of (inverse) diffusion is actually a measure of relative depth. Because of ill-posedness we propose a graph-cuts based method for inferring the depth in the scene using the amount of diffusion as a data likelihood and a smoothness condition on the depth in the scene. The method is verified experimentally on a varied set of test cases.

In the area of depth estimation from images an interesting approach has been structure recovery from defocus cue. Towards this end, there have been a number of approaches [4,6]. Here we propose a technique to estimate the regularized depth from defocus using diffusion. The coefficient of the diffusion equation is modeled using a pair-wise Markov random field (MRF) ensuring spatial regularization to enhance the robustness of the depth estimated. This framework is solved efficiently using a graph-cuts based techniques. The MRF representation is enhanced by incorporating a smoothness prior that is obtained from a graph based segmentation of the input images. The method is demonstrated on a number of data sets and its performance is compared with state of the art techniques.

We consider the problem of depth estimation from multiple images based on the defocus cue. For a Gaussian defocus blur, the observations can be shown to be the solution of a deterministic but inhomogeneous diffusion process. However, the diffusion process does not sufficiently address the case in which the Gaussian kernel is deformed. This deformation happens due to several factors like self-occlusion, possible aberrations and imperfections in the aperture. These issues can be solved by incorporating a stochastic perturbation into the heat diffusion process. The resultant flow is that of an inhomogeneous heat diffusion perturbed by a stochastic curvature driven motion. The depth in the scene is estimated from the coefficient of the stochastic heat equation without actually knowing the departure from the Gaussian assumption. Further, the proposed method also takes into account the non-convex nature of the diffusion process. The method provides a strong theoretical framework for handling the depth from defocus problem. 1

Blind restoration of blurred images is a classical ill-posed problem. There has been considerable interest in the use of partial differential equations to solve this problem. The blurring of an image has traditionally been modeled by Witkin [10] and Koenderink [4] by the heat equation. This has been the basis of the Gaussian scale space. However, a similar theoretical formulation has not been possible for deblurring of images due to the ill-posed nature of the reverse heat equation. Here we consider the stabilization of the reverse heat equation. We do this by damping the distortion along the edges by adding a normal component of the heat equation in the forward direction. We use a stopping criterion based on the divergence of the curvature in the resulting reverse heat flow. The resulting stabilized reverse heat flow makes it possible to solve the challenging problem of blind space varying deconvolution. The method is justified by a varied set of experimental results.

One of the classic problems in low level vision is image restoration. An important contribution toward this effort has been the development of shock filters by Osher and Rudin (1990). It performs image deblurring using hyperbolic partial differential equations. In this paper we relate the notion of cluster separation from the field of pattern recognition to the shock filter formulation. A kind of shock filter is proposed based on the idea of gradient based separation of clusters. The proposed formulation is general enough as it can allow various models of density functions in the cluster separation process. The efficacy of the method is demonstrated through various examples.

We propose an image retrieval scheme based on projectively invariant features. Since cross-ratio is the fundamental invariant feature under projective transformations for points, we use that as the basic feature parameter. We compute the cross-ratios of point sets in quadruplets and a discrete representation of the distribution of the cross-ratio is obtained from the computed values. The distribution is used as the feature for retrieval purposes. The method is very effective in retrieving images, like buildings, having similar planar 3D structures.