Nanjing University of Science and Technology

About: Nanjing University of Science and Technology is a education organization based out in Nanjing, China. It is known for research contribution in the topics: Control theory & Catalysis. The organization has 31581 authors who have published 36390 publications receiving 525474 citations. The organization is also known as: Nánjīng Lǐgōng Dàxué & Nánlǐgōng.

Rethinking the Bottom-Up Framework for Query-Based Video Localization

Abstract: In this paper, we focus on the task query-based video localization, i.e., localizing a query in a long and untrimmed video. The prevailing solutions for this problem can be grouped into two categories: i) Top-down approach: It pre-cuts the video into a set of moment candidates, then it does classification and regression for each candidate; ii) Bottom-up approach: It injects the whole query content into each video frame, then it predicts the probabilities of each frame as a ground truth segment boundary (i.e., start or end). Both two frameworks have respective shortcomings: the top-down models suffer from heavy computations and they are sensitive to the heuristic rules, while the performance of bottom-up models is behind the performance of top-down counterpart thus far. However, we argue that the performance of bottom-up framework is severely underestimated by current unreasonable designs, including both the backbone and head network. To this end, we design a novel bottom-up model: Graph-FPN with Dense Predictions (GDP). For the backbone, GDP firstly generates a frame feature pyramid to capture multi-level semantics, then it utilizes graph convolution to encode the plentiful scene relationships, which incidentally mitigates the semantic gaps in the multi-scale feature pyramid. For the head network, GDP regards all frames falling in the ground truth segment as the foreground, and each foreground frame regresses the unique distances from its location to bi-directional boundaries. Extensive experiments on two challenging query-based video localization tasks (natural language video localization and video relocalization), involving four challenging benchmarks (TACoS, Charades-STA, ActivityNet Captions, and Activity-VRL), have shown that GDP surpasses the state-of-the-art top-down models.

Robust subspace segmentation via low-rank representation.

Abstract: Recently the low-rank representation (LRR) has been successfully used in exploring the multiple subspace structures of data. It assumes that the observed data is drawn from several low-rank subspaces and sometimes contaminated by outliers and occlusions. However, the noise (low-rank representation residual) is assumed to be sparse, which is generally characterized by minimizing the l1 -norm of the residual. This actually assumes that the residual follows the Laplacian distribution. The Laplacian assumption, however, may not be accurate enough to describe various noises in real scenarios. In this paper, we propose a new framework, termed robust low-rank representation, by considering the low-rank representation as a low-rank constrained estimation for the errors in the observed data. This framework aims to find the maximum likelihood estimation solution of the low-rank representation residuals. We present an efficient iteratively reweighted inexact augmented Lagrange multiplier algorithm to solve the new problem. Extensive experimental results show that our framework is more robust to various noises (illumination, occlusion, etc) than LRR, and also outperforms other state-of-the-art methods.