Search

Your search keyword '"Liu, Zixin"' showing total 638 results
Start Over Author "Liu, Zixin"
638 results on '"Liu, Zixin"'

1. OccFusion: Depth Estimation Free Multi-sensor Fusion for 3D Occupancy Prediction

Author

Zhang, Ji, Ding, Yiran, and Liu, Zixin

Subjects
Computer Science - Computer Vision and Pattern Recognition
Abstract

3D occupancy prediction based on multi-sensor fusion,crucial for a reliable autonomous driving system, enables fine-grained understanding of 3D scenes. Previous fusion-based 3D occupancy predictions relied on depth estimation for processing 2D image features. However, depth estimation is an ill-posed problem, hindering the accuracy and robustness of these methods. Furthermore, fine-grained occupancy prediction demands extensive computational resources. To address these issues, we propose OccFusion, a depth estimation free multi-modal fusion framework. Additionally, we introduce a generalizable active training method and an active decoder that can be applied to any occupancy prediction model, with the potential to enhance their performance. Experiments conducted on nuScenes-Occupancy and nuScenes-Occ3D demonstrate our framework's superior performance. Detailed ablation studies highlight the effectiveness of each proposed method.

Published
2024

2. Using fine-tuning and min lookahead beam search to improve Whisper

Author

Do, Andrea, Brown, Oscar, Wang, Zhengjie, Mathew, Nikhil, Liu, Zixin, Ahmed, Jawwad, and Yu, Cheng

Subjects
Electrical Engineering and Systems Science - Audio and Speech Processing, Computer Science - Computation and Language, Computer Science - Machine Learning, Computer Science - Sound
Abstract

The performance of Whisper in low-resource languages is still far from perfect. In addition to a lack of training data on low-resource languages, we identify some limitations in the beam search algorithm used in Whisper. To address these issues, we fine-tune Whisper on additional data and propose an improved decoding algorithm. On the Vietnamese language, fine-tuning Whisper-Tiny with LoRA leads to an improvement of 38.49 in WER over the zero-shot Whisper-Tiny setting which is a further reduction of 1.45 compared to full-parameter fine-tuning. Additionally, by using Filter-Ends and Min Lookahead decoding algorithms, the WER reduces by 2.26 on average over a range of languages compared to standard beam search. These results generalise to larger Whisper model sizes. We also prove a theorem that Min Lookahead outperforms the standard beam search algorithm used in Whisper., Comment: 8 pages, submitted to IEEE ICASSP 2024

Published
2023

23. Main-Aftershock Modeling According to Measured Records

Author

Liu, Zixin, Jiang, Yunmu, Liu, Zhangjun, Ruan, Xinxin, di Prisco, Marco, Series Editor, Chen, Sheng-Hong, Series Editor, Vayas, Ioannis, Series Editor, Kumar Shukla, Sanjay, Series Editor, Sharma, Anuj, Series Editor, Kumar, Nagesh, Series Editor, Wang, Chien Ming, Series Editor, Guo, Wei, editor, and Qian, Kai, editor

Published
2023
Full Text
View/download PDF

39. Preparation of cashew nut shell liquid based hypercrosslinked polymers and VOCs adsorption performance

Author

LIU Zixin, LIU Huan, YU Haoran, WANG Yan, and WANG Jianying

Subjects
hypercrosslinked polymer, preparation, adsorption, vocs, regeneration, Materials of engineering and construction. Mechanics of materials, TA401-492
Abstract

Hypercrosslinked polymers(HCPs) were prepared by cashew nut shell liquid(CNSL), an agricultural by-product, as a raw material, using external cross-linker method and solvent cross-linker method, and named F-HCP and C-HCP, respectively. The adsorption properties of volatile organic compounds(VOCs) on F-HCP and C-HCP were investigated. The prepared adsorbent was characterized by infrared spectroscopy(FTIR), thermogravimetric analysis(TGA), N2 adsorption and scanning electron microscopy(SEM). The adsorption properties of six kinds of VOCs on F-HCP and C-HCP were investigated by static adsorption method. The o-xylene was used as a typical VOC for dynamic adsorption, and the influence of inlet concentration on adsorption was investigated. The results show that F-HCP has higher static adsorption performance than C-HCP, and the static adsorption capacities of o-xylene, toluene, cyclohexane, n-hexane, methyl ethyl ketone, and acetone on F-HCP are 930, 801, 300, 214, 203 mg·g-1and 164 mg·g-1, respectively. At the inlet concentration of 2171 mg·m-3 for o-xylene, the dynamic adsorption capacity of F-HCP and C-HCP for o-xylene is 149 mg·g-1 and 146 mg·g-1, respectively. The equilibrium adsorption capacity increases with the increase of the inlet gas concentration. The cyclic regeneration adsorption experiments indicate that the adsorption capacity of C-HCP after five cycles of adsorption is 93.9% of the first adsorption capacity.

Published
2023
Full Text
View/download PDF

46. Programmable Neural Network Trojan for Pre-Trained Feature Extractor

Author

Ji, Yu, Liu, Zixin, Hu, Xing, Wang, Peiqi, and Zhang, Youhui

Subjects
Computer Science - Cryptography and Security, Computer Science - Computer Vision and Pattern Recognition, Computer Science - Neural and Evolutionary Computing
Abstract

Neural network (NN) trojaning attack is an emerging and important attack model that can broadly damage the system deployed with NN models. Existing studies have explored the outsourced training attack scenario and transfer learning attack scenario in some small datasets for specific domains, with limited numbers of fixed target classes. In this paper, we propose a more powerful trojaning attack method for both outsourced training attack and transfer learning attack, which outperforms existing studies in the capability, generality, and stealthiness. First, The attack is programmable that the malicious misclassification target is not fixed and can be generated on demand even after the victim's deployment. Second, our trojan attack is not limited in a small domain; one trojaned model on a large-scale dataset can affect applications of different domains that reuse its general features. Thirdly, our trojan design is hard to be detected or eliminated even if the victims fine-tune the whole model.

Published
2019

Catalog

Books, media, physical & digital resources
See catalog results