Ning Zhang

Jiayuan Wang, Q. M. Jonathan Wu, Ning Zhang et al.

Autonomous driving systems rely on panoptic perception to jointly handle object detection, drivable area segmentation, and lane line segmentation. Although multi-task learning is an effective way to integrate these tasks, its increasing model parameters and complexity make deployment on on-board devices difficult. To address this challenge, we propose a multi-task model compression framework that combines task-aware safe pruning with feature-level knowledge distillation. Our safe pruning strategy integrates Taylor-based channel importance with gradient conflict penalty to keep important channels while removing redundant and conflicting channels. To mitigate performance degradation after pruning, we further design a task head-agnostic distillation method that transfers intermediate backbone and encoder features from a teacher to a student model as guidance. Experiments on the BDD100K dataset demonstrate that our compressed model achieves a 32.7% reduction in parameters while segmentation performance shows negligible accuracy loss and only a minor decrease in detection (-1.2% for Recall and -1.8% for mAP50) compared to the teacher. The compressed model still runs at 32.7 FPS in real-time. These results show that combining pruning and knowledge distillation provides an effective compression solution for multi-task panoptic perception.

Wenxuan Liu, Tianyao He, Chen Gong et al.

Music plagiarism detection is gaining more and more attention due to the popularity of music production and society's emphasis on intellectual property. We aim to find fine-grained plagiarism in music pairs since conventional methods are coarse-grained and cannot match real-life scenarios. Considering that there is no sizeable dataset designed for the music plagiarism task, we establish a large-scale simulated dataset, named Music Plagiarism Detection Dataset (MPD-Set) under the guidance and expertise of renowned researchers from national-level professional institutions in the field of music. MPD-Set considers diverse music plagiarism cases found in real life from the melodic, rhythmic, and tonal levels respectively. Further, we establish a Real-life Dataset for evaluation, where all plagiarism pairs are real cases. To detect the fine-grained plagiarism pairs effectively, we propose a graph-based method called Bipatite Melody Matching Detector (BMM-Det), which formulates the problem as a max matching problem in the bipartite graph. Experimental results on both the simulated and Real-life Datasets demonstrate that BMM-Det outperforms the existing plagiarism detection methods, and is robust to common plagiarism cases like transpositions, pitch shifts, duration variance, and melody change. Datasets and source code are open-sourced at https://github.com/xuan301/BMMDet_MPDSet.

Ning Zhang, Susan Francis, Rayaz Malik et al.

Semantic image segmentation is one of the most important tasks in medical image analysis. Most state-of-the-art deep learning methods require a large number of accurately annotated examples for model training. However, accurate annotation is difficult to obtain especially in medical applications. In this paper, we propose a spatially constrained deep convolutional neural network (DCNN) to achieve smooth and robust image segmentation using inaccurately annotated labels for training. In our proposed method, image segmentation is formulated as a graph optimization problem that is solved by a DCNN model learning process. The cost function to be optimized consists of a unary term that is calculated by cross entropy measurement and a pairwise term that is based on enforcing a local label consistency. The proposed method has been evaluated based on corneal confocal microscopic (CCM) images for nerve fiber segmentation, where accurate annotations are extremely difficult to be obtained. Based on both the quantitative result of a synthetic dataset and qualitative assessment of a real dataset, the proposed method has achieved superior performance in producing high quality segmentation results even with inaccurate labels for training.

Ning Zhang, Junchi Yan, Yuchen Zhou

Separating audio mixtures into individual instrument tracks has been a long standing challenging task. We introduce a novel weakly supervised audio source separation approach based on deep adversarial learning. Specifically, our loss function adopts the Wasserstein distance which directly measures the distribution distance between the separated sources and the real sources for each individual source. Moreover, a global regularization term is added to fulfill the spectrum energy preservation property regardless separation. Unlike state-of-the-art weakly supervised models which often involve deliberately devised constraints or careful model selection, our approach need little prior model specification on the data, and can be straightforwardly learned in an end-to-end fashion. We show that the proposed method performs competitively on public benchmark against state-of-the-art weakly supervised methods.

Ning Zhang, Evan Shelhamer, Yang Gao et al.

Pose variation and subtle differences in appearance are key challenges to fine-grained classification. While deep networks have markedly improved general recognition, many approaches to fine-grained recognition rely on anchoring networks to parts for better accuracy. Identifying parts to find correspondence discounts pose variation so that features can be tuned to appearance. To this end previous methods have examined how to find parts and extract pose-normalized features. These methods have generally separated fine-grained recognition into stages which first localize parts using hand-engineered and coarsely-localized proposal features, and then separately learn deep descriptors centered on inferred part positions. We unify these steps in an end-to-end trainable network supervised by keypoint locations and class labels that localizes parts by a fully convolutional network to focus the learning of feature representations for the fine-grained classification task. Experiments on the popular CUB200 dataset show that our method is state-of-the-art and suggest a continuing role for strong supervision.

Jonathan Long, Ning Zhang, Trevor Darrell

Convolutional neural nets (convnets) trained from massive labeled datasets have substantially improved the state-of-the-art in image classification and object detection. However, visual understanding requires establishing correspondence on a finer level than object category. Given their large pooling regions and training from whole-image labels, it is not clear that convnets derive their success from an accurate correspondence model which could be used for precise localization. In this paper, we study the effectiveness of convnet activation features for tasks requiring correspondence. We present evidence that convnet features localize at a much finer scale than their receptive field sizes, that they can be used to perform intraclass alignment as well as conventional hand-engineered features, and that they outperform conventional features in keypoint prediction on objects from PASCAL VOC 2011.