Abstract

Table structure recognition is necessary for a compre- hensive understanding of documents. Tables in unstruc- tured business documents are tough to parse due to the high diversity of layouts, varying alignments of contents, and the presence of empty cells. The problem is particularly dif- ficult because of challenges in identifying individual cells using visual or linguistic contexts or both. Accurate detec- tion of table cells (including empty cells) simplifies struc- ture extraction and hence, it becomes the prime focus of our work. We propose a novel object-detection-based deep model that captures the inherent alignments of cells within tables and is fine-tuned for fast optimization. Despite ac- curate detection of cells, recognizing structures for dense tables may still be challenging because of difficulties in cap- turing long-range row/column dependencies in presence of multi-row/column spanning cells. Therefore, we also aim to improve structure recognition by deducing a novel rec- tilinear graph-based formulation. From a semantics per- spective, we highlight the significance of empty cells in a table. To take these cells into account, we suggest an en- hancement to a popular evaluation criterion. Finally, we introduce a modestly sized evaluation dataset with an anno- tation style inspired by human cognition to encourage new approaches to the problem. Our framework improves the previous state-of-the-art performance by a 2.7% average F1-score on benchmark datasets.

Abstract

Advancements in adaptive object detection can lead to tremendous improvements in applications like autonomous navigation, as they alleviate the distributional shifts along the detection pipeline. Prior works adopt adversarial learn- ing to align image features at global and local levels, yet the instance-specific misalignment persists. Also, adaptive object detection remains challenging due to visual diver- sity in background scenes and intricate combinations of ob- jects. Motivated by structural importance, we aim to attend prominent instance-specific regions, overcoming the feature misalignment issue. We propose a novel resIduaL seLf- attentive featUre alignMEnt (ILLUME) method for adaptive object detection. ILLUME comprises Self-Attention Fea- ture Map (SAFM) module that enhances structural atten- tion to object-related regions and thereby generates domain invariant features. Our approach significantly reduces the domain distance with the improved feature alignment of the instances. Qualitative results demonstrate the ability of ILLUME to attend important object instances required for alignment. Experimental results on several benchmark datasets show that our method outperforms the existing state-of-the-art approaches.

Abstract

Player detection lays the foundation for many applica- tions in the field of sports analytics including player recog- nition, player tracking, and activity detection. In this work, we study player detection in continuous long shot broad- cast videos. Broadcast match videos are easy to obtain, and detection on these videos is much more challenging. We propose a transductive approach for player detection that treats it as a domain adaptation problem. We show that instance-level domain labels are significant for sufficient adaptation in the case of soccer broadcast videos. An ef- ficient multi-model greedy labelling scheme based on visual features is proposed to annotate domain labels on bounding box predictions made by our inductive model. We use reli- able instances from the inductive model inferences to train a transductive copy of the model. We create and release a fully annotated player detection dataset comprising soccer broadcast videos from the FIFA 2018 World Cup matches to evaluate our method. Our method shows significant im- provements in player detection to the baseline and existing state-of-the-art methods on our dataset. We show, on aver- age, a 16 point improvement in mAP for soccer broadcast videos by annotating domain labels for around a 100 sam- ples per video.

Abstract

Systemic lupus erythematosus (SLE) is an autoimmune disease in which the immune system of the patient starts attacking healthy tissues of the body. Lupus Nephritis (LN) refers to the inflammation of kidney tissues resulting in renal failure due to these attacks. The International Society of Nephrology/Renal Pathology Society (ISN/RPS) has released a classification system based on various patterns observed during renal injury in SLE. Traditional methods require meticulous pathological assessment of the renal biopsy and are time-consuming. Recently, computational techniques have helped to alleviate this issue by using virtual microscopy or Whole Slide Imaging (WSI). With the use of deep learning and modern computer vision techniques, we propose a pipeline that is able to automate the process of 1) detection of various glomeruli patterns present in these whole slide images and 2) classification of each image using the extracted glomeruli features.

Abstract

The identification of plant species by looking at their leaves, flowers, and seeds is a crucial component in the conservation of endangered plants. Traditional identification methods are manual and time consuming and require domain knowledge to operate. Owing to an increased interest in the automated plant identification system, we propose one that utilizes modern convolutional neural network architectures. This approach helps in the recognition of leaf images and can be integrated into mobile platforms like smartphones. Such a system can also be employed in aiding plant-related education, promoting ecotourism, and creating a digital heritage for plant species, among many others. Our proposed solution achieves a state-of-the-art performance for plant classification in the wild. An exhaustive set of experiments are performed to classify 112 species of plants from the challenging Indic-Leaf dataset. The best-performing model gives Top 1 precision of 90.08 and Top 5 precision of 96.90. We discuss and elaborate on our crowdsourcing web application that is used to collect and regulate data. We explain how the automated plant identification system can be integrated into a smartphone by detailing the flow of our mobile application.

Abstract

As Deep Neural Network models for face processing tasks approach human-like performance, their deployment in critical applications such as law enforcement and access control has seen an upswing, where any failure may have far-reaching consequences. We need methods to build trust in deployed systems by making their working as transparent as possible. Existing visualization algorithms are designed for object recognition and do not give insightful results when applied‘ to the face domain. In this work, we present ‘Canonical Saliency Maps’, a new method which highlights relevant facial areas by projecting saliency maps onto a canonical face model. We present two kinds of Canonical Saliency Maps: image-level maps and model-level maps. Image-level maps highlight facial features responsible for the decision made by a deep face model on a given image, thus helping to understand how a DNN made a prediction on the image. Model-level maps provide an understanding of what the entire DNN model focuses on in each task, and thus can be used to detect biases in the model. Our qualitative and quantitative results show the usefulness of the proposed canonical saliency maps, which can be used on any deep face model regardless of the architecture.

Abstract

Few-shot learners aim to recognize new categories given only a small number of training samples. The core challenge is to avoid overfitting to the limited data while ensuring good generalization to novel classes. Existing literature makes use of vast amounts of annotated data by simply shifting the label requirement from novel classes to base classes. Since data annotation is time-consuming and costly, reducing the label requirement even further is an important goal. To that end, our paper presents a more challenging few-shot setting with almost no class label access. By leveraging self-supervision to learn image representations and similarity for classification at test time, we achieve competitive baselines while using almost zero (0-5) class labels. Compared to existing state-of-the-art approaches which use 60,000 labels, this is a four orders of magnitude (10,000 times) difference. This work is a step towards developing few-shot learning methods that do not depend on annotated data. Our code is publicly released at https: //github.com/adbugger/FewShot. 3

Abstract

Handwritten documents from communities like cultural heritage, judiciary, and modern journals remain largely unexplored even today. To a great extent, this is due to the lack of retrieval tools for such unlabeled document collections. This work considers such collections and presents a simple, robust retrieval framework for easy information access. We achieve retrieval on unlabeled novel collections through invariant features learned for handwritten text. These feature representations enable zero-shot retrieval for novel queries on unlabeled collections. We improve the framework further by supporting search via text and exemplar queries. Four new collections written in English, Malayalam, and Bengali are used to evaluate our text retrieval framework. These collections comprise 2957 handwritten pages and over 300K words. We report promising results on these collections, despite the zero-shot constraint and huge collection size. Our framework allows the addition of new collections without any need for specific finetuning or labeling. Finally, we also present a demonstration of the retrieval framework. [Project Page] Keywords: Document retrieval· Keyword Spotting· Zero-shot retrieval

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.