Abstract

Advances in deep learning have been instrumental in enhancing the performance of face verification systems. Despite their ability to attain high accuracy, most of these systems fail to provide interpretations of their decisions. With the increased demands in making deep learning models more interpretable, numerous post-hoc methods have been proposed to probe the workings of these systems. Yet, the quest for face verification systems that inherently provide interpretations still remains largely unexplored. Additionally, most of the existing face recognition models are highly susceptible to adversarial attacks. In this work, we propose a face verification system which addresses the issue of interpretability by employing modular neural networks. In this, representations for each individual facial parts such as nose, mouth, eyes etc. are learned separately. We also show that our method is significantly more resistant to adversarial attacks, thereby addressing another crucial weakness concerning deep learning models.

Abstract

Speech is a natural mode of communication among all beings. India is a densely populated country, and people are diverse throughout the globe. The spoken language is the medium of instruction to interact among the people. The majority of Indian languages are spoken globally. The unavailability of larger volumes of transcribed and annotated speech data is often a hurdle for building reliable speech recognition (ASR) systems for Indian languages. Crowdsourcing strategies are effective in collaboratively collecting speech data resources. This paper describes the experience of large-scale speech data collection for the Telugu language through mobile and web-based applications. With this crowd contributed speech, the performance of the baseline ASR system is shown for clean speech. ASR performance for pink and white noises is also compared for various deep neural network (DNN) based acoustic models. The details regarding the usage of frameworks and their challenges during their implementation are part of this paper. The framework adopted for collecting the speech data is rapid, cost-saving, and offers the advantage of extending it to all the other Indian languages. Index Terms—ASR, Crowd-sourced, TDNN, GMM, SGMM

Abstract

Reinforcement learning (RL) has made significant progress in both abstract and real-world domains, but the majority of state-of-the-art algorithms deal only with monotonic actions. However, some applications require agents to reason over different types of actions. Our application simulates reactionbased molecule generation, used as part of the drug discovery pipeline, and includes both uni-molecular and bi-molecular reactions. This paper introduces a novel framework, towered actor critic (TAC), to handle multiple action types. The TAC framework is general in that it is designed to be combined with any existing RL algorithms for continuous action space. We combine it with TD3 to empirically obtain significantly better results than existing methods in the drug discovery setting. TAC is also applied to RL benchmarks in OpenAI Gym and results show that our framework can improve, or at least does not hurt, performance relative to standard TD3.

Abstract

Object detection, semantic segmentation, and instance seg-mentation form the bases for many computer vision tasks in autonomous driving. The complexity of these tasks increases as we shift from object detection to instance segmentation. The state-of-the-art models are eval-uated on standard datasets such as pascal-voc and ms-cococ, which do not consider the dynamics of road scenes. Driving datasets such as Cityscapes and Berkeley Deep Drive (bdd) are captured in a structured environment with better road markings and fewer variations in the ap-pearance of objects and background. However, the same does not hold for Indian roads. The Indian Driving Dataset (idd) is captured in un-structured driving scenarios and is highly challenging for a model due toits diversity. This work presents a comprehensive evaluation of state-of-the-art models on object detection, semantic segmentation, and instance segmentation on-road scene datasets. We present our analyses and com- pare their quantitative and qualitative performance on structured driv-ing datasets (Cityscapes and bdd) and the unstructured driving dataset (idd); understanding the behavior on these datasets helps in addressing various practical issues and helps in creating real-life applications.

Abstract

Object detection is a key component in autonomous navigation systems that enables localization and classification of the objects in a road scene. Existing object detection methods are trained and inferred on a fixed number of known classes present in road scenes. However, in real-world or open-world road scenes, while inference, we come across unknown objects that the detection model hasn’t seen while training. Hence, we propose O pen Wo rld Object Detection on Road Scenes (ORDER) to address the aforementioned problem for road scenes. Firstly, we introduce Feature-Mix to improve the unknown object detection capabilities of an object detector. Feature-Mix widens the gap between known and unknown classes in latent feature space that helps improve the unknown object detection. Next, we identify that the road scene dataset compared to generic object dataset contains a significant proportion of small objects and has higher intra-class bounding box scale variations, making it challenging to detect the known and unknown objects. We propose a novel loss: Focal regression loss that collectively addresses the problem of small object detection and intra-class bounding box by penalizing more the small bounding boxes and dynamically changing the loss according to object size. Further, the detection of small objects is improved by curriculum learning. Finally, we present an extensive evaluation on two road scene datasets: BDD and IDD. Our experimental evaluations on BDD and IDD shows consistent improvement over the current state-of-the-art method. We believe that this work will lay the foundation for real-world object detection for road scenes.

Abstract

Droughts are recognized as a natural disaster that is caused by extreme and continuous shortage of precipitation. Drought indices assist in a number of tasks, including its early warning and monitoring by computing severity levels and proclaiming the start and end of drought. Various drought indices were formulated for the forecasting and prediction of spatiotemporal drought characteristics using various hydrological variables, such as precipitation, evapotranspiration, runoff, soil moisture content, etc. Due to anthropogenic global warming and increase of temperature, evapotranspiration based drought indices have become interest in recent years in the drought assessment. This paper attempt to provide more information on drought indices which incorporates Evapotranspiration. The study used Standardized Precipitation Evapotranspiration index (SPEI), based on evapotranspiration to understand the drought variability at various time scales. This study adopted Hargreaves model to calculate Potential Evapotranspiration. SPEI index can also be used to study the wet and dry periods including Evapotranspiration along with precipitation. The study used SPEI to understand the dry and wet years over an urban semi-arid region, Hyderabad, capital and biggest city of the southern Indian state of Telangana for the year 1965 to 2015. The years 1965, 1966, 1972, 1973, 1985, 1993 and 2012 were noted as dry years with SPEI values as -1.35,-1.06,-1.43,-1.31,-1.05, -1.22 and -1.51 sequentially and year 2006 as severe wet year with SPEI value as +1.65. The characterization of dry and wet years as demonstrated in the present study will enhance the better urban water resources management.

Abstract

The nationwide lockdown was imposed over In- dia from 25 March to 31 May 2020 with varied relax- ations from phase I to phase IV to contain the spread of COVID-19. Thus, emissions from industrial and transport sectors were halted during lockdown (LD), which has re-sulted in a significant reduction of anthropogenic pollutants. The first two lockdown phases were strictly implemented (phase I and phase II) and hence were considered to be total lockdown (TLD) in this study. Satellite-based tropospheric columnar nitrogen dioxide (TCN) from the years 2015 to 2020, tropospheric columnar carbon monoxide (TCC) dur-ing 2019/20, and aerosol optical depth (AOD550) from the years 2014 to 2020 during phase I and phase II LD and pre- LD periods were investigated with observations from Aura OMI, Sentinel-5P TROPOMI, and Aqua and Terra MODIS. To quantify lockdown-induced changes in TCN, TCC, and AOD550, detailed statistical analysis was performed on de-trended data using the Student paired statistical t test. Re-sults indicate that mean TCN levels over India showed a dipof 18 % compared to the previous year and also against the 5-year mean TCN levels during the phase I lockdown, which was found to be statistically significant (p value < 0.05) against the respective period. Furthermore, drastic changes in TCN levels were observed over hotspots, namely eastern region and urban cities. For example, there was a sharp de-crease of 62 % and 54 % in TCN levels compared to 2019 and against 5-year mean TCN levels over New Delhi with a p value of 0.0002 (which is statistically significant) dur-ing total LD. The TCC levels were high in the northeast (NE) region during the phase I LD period, which is mainly attributed to the active fire counts in this region. However, lower TCC levels are observed in the same region due to the diminished fire counts during phase II. Further, AOD550 is reduced over the country by ∼ 16 % (Aqua and Terra) from the 6-year (2014–2019) mean AOD550 levels, with a signifi-cant reduction (Aqua MODIS 28 %) observed over the Indo- Gangetic Plain (IGP) region with a p value of  0.05. How-ever, an increase in AOD550 levels (25 % for Terra MODIS, 15 % for Aqua MODIS) was also observed over central India during LD compared to the preceding year and found signif-icant with a p value of 0.03. This study also reports the rate of change of TCN levels and AOD550 along with statistical metrics during the LD period.

Abstract

How bad is my wound? How fast will the wound heal? Do I need to get hospitalized? Questions like these are critical for wound assessment, but challenging to answer. Given a wound image and patient attributes, our goal is to build models for two wound assessment tasks: (1) predicting if the patient needs hospitalization for the wound to heal, and (2) estimating wound progression, i.e., weeks to heal. The problem is challenging because wound progression and hospitalization risk depend on multiple factors that need to be inferred automatically from the given wound image. There exists no work which performs a rigorous study of wound assessment tasks considering multiple wound attributes inferred using a large dataset of wound images. We present HealTech, a two-stage wound assessment solution. The first stage predicts various wound attributes (like ulcer type, location, stage, etc.) from wound images, using deep neural networks. The second stage predicts (1) whether the wound would heal (using conventional in-house treatment) or not (needs hospitalization), and (2) the number of weeks to heal, using an evolutionary algorithm based stacked Light Gradient Boosted Machines (LGBM) model. On a large dataset of 125711 wound images, HealTech achieves a recall of 83 and a precision of 92 for wounds with the risk of hospitalization. For wounds that can be healed without hospitalization, precision and recall are as high as 99. Our wound progression model provides a mean absolute error of 3.3 weeks.

Abstract

Performing sound source separation and visual object seg- mentation jointly in naturally occurring videos is a notori- ously difficult task, especially in the absence of annotated data. In this study, we leverage the concurrency between audio and visual modalities in an attempt to solve the joint audio-visual segmentation problem in a self-supervised man- ner. Human beings interact with the physical world through a few sensory systems such as vision, auditory, movement, etc. The usefulness of the interplay of such systems lies in the con- cept of degeneracy [1]. It tells us that the cross-modal signals can educate each other without the presence of an external supervisor. In this work, we efficiently exploit this fact that learning from one modality inherently helps to find patterns in others by introducing a novel audio-visual fusion technique. Also, to the best of our knowledge, we are the first to ad- dress the partially occluded sound source segmentation task. Our study shows that the proposed model significantly out- performs existing state-of-the-art methods in both visual and audio source separation tasks

Abstract

Estimation of time delay from the received broadband signals like speech, collected at two or more spatially distributed microphones, has many applications. Methods like the cross-correlation of the signals directly and generalized crosscorrelation based methods (GCC and GCC-PHAT) have been used for several years to estimate the time delay. Performance of these methods degrades due to noise, multi-path reflections, and reverberation in a practical environment, like a live room. The estimated time delay is usually robust due to the averaging effect of the delay obtained over several frames in an utterance of a few seconds. The robustness is affected if the varying time delay of a moving speaker is desired. A smaller duration for averaging results in errors in the estimation of the time delay, and a longer duration for averaging results in loss of accuracy. Since the single frequency filtering (SFF) based analysis provides an estimation of the instantaneous time delay, it is possible to study the trade off between accuracy and robustness. This paper examines this trade-off in determining the number of stationary speakers from mixed signals and in tracking a speaker moving along a straight line path and along a circular path. The results are illustrated for actual data collected in a live room.