Referring expression (RE) for video domain describes the video using a natural language expression. Relational RE comprehension in a video domain localizes an object in relation to a distinguishing context object. Unlike object grounding in videos using REs, not much work has been done in videos using relational REs. In this paper, we focus on (1) relational RE comprehension for videos, and (2) demonstrating the significance of attention for the task. We propose a novel modular network based approach for relational RE comprehension in highly ambiguous settings for videos. We show the significance of the language attention in modular approach by: (1) Using two different networks, i.e., modATN consisting of attention mechanism, visual modules, and a natural language expression input, and modSTR consisting of visual modules, and structured input (subject, subject adjective, object, object adjective, action, relation); (2) Introducing a new dataset having structured RE for relation RE comprehension task in modSTR. Finally, we propose an optimised modular network that outperforms and shows significant improvements over the baseline networks.

Proteins/peptides have shown to be promising therapeutic agents for a variety of diseases. However, toxicity is one of the obstacles in protein/peptide-based therapy. The current study describes a web-based tool, ToxinPred2, developed for predicting the toxicity of proteins. This is an update of ToxinPred developed mainly for predicting toxicity of peptides and small proteins. The method has been trained, tested and evaluated on three datasets curated from the recent release of the SwissProt. To provide unbiased evaluation, we performed internal validation on 80% of the data and external validation on the remaining 20% of data. We have implemented the following techniques for predicting protein toxicity; (i) Basic Local Alignment Search Tool-based similarity, (ii) Motif-EmeRging and with Classes-Identification-based motif search and (iii) Prediction models. Similarity and motif-based techniques achieved a high probability of correct prediction with poor sensitivity/coverage, whereas models based on machine-learning techniques achieved balance sensitivity and specificity with reasonably high accuracy. Finally, we developed a hybrid method that combined all three approaches and achieved a maximum area under receiver operating characteristic curve around 0.99 with Matthews correlation coefficient 0.91 on the validation dataset. In addition, we developed models on alternate and realistic datasets. The best machine learning models have been implemented in the web server named 'ToxinPred2', which is available at https://webs.iiitd.edu.in/raghava/toxinpred2/ and a standalone version at https://github.com/raghavagps/toxinpred2. This is a general method developed for predicting the toxicity of proteins regardless of their source of origin.

The state-of-the-art object detection and image classification methods can perform impressively on more than 9k classes. In contrast, the number of classes in semantic segmentation datasets is relatively limited. This is not surprising when the restrictions caused by the lack of labeled data and high computation demand for segmentation are considered. In this paper, we propose a novel training methodology to train and scale the existing semantic segmentation models for a large number of semantic classes without increasing the memory overhead. In our embedding-based scalable segmentation approach, we reduce the space complexity of the segmentation model’s output from O(C) to O(1), propose an approximation method for ground-truth class probability, and use it to compute cross-entropy loss. The proposed approach is general and can be adopted by any state-of-the-art segmentation model to gracefully scale it for any number of semantic classes with only one GPU. Our approach achieves similar, and in some cases, even better mIoU for Cityscapes, Pascal VOC, ADE20k, COCO-Stuff10k datasets when adopted to DeeplabV3+ model with different backbones. We demonstrate a clear benefit of our approach on a dataset with 1284 classes, bootstrapped from LVIS and COCO annotations, with almost three times better mIoU than the DeeplabV3+. Our source code is available at: https://github.com/shipra25jain/ESSNet.