Vehicle Pose Transform by Generative Adversarial Network for Re-Identification
碩士 === 國立中正大學 === 資訊工程研究所 === 107 === Vehicle Re-identification (Re-ID) is an major task that seeks a query vehicleimage from the gallery image dataset and it has the huge potential to contribute tothe intelligent video surveillance. However, the pose variation of vehicle images isone of the key ch...
| Main Authors: | , |
|---|---|
| Other Authors: | |
| Format: | Others |
| Language: | en_US |
| Published: |
2019
|
| Online Access: | http://ndltd.ncl.edu.tw/handle/ev7rhw |
| Summary: | 碩士 === 國立中正大學 === 資訊工程研究所 === 107 === Vehicle Re-identification (Re-ID) is an major task that seeks a query vehicleimage from the gallery image dataset and it has the huge potential to contribute tothe intelligent video surveillance. However, the pose variation of vehicle images isone of the key challenges. Same vehicle identities with different viewpoint usuallyhave large discrepancy. In this paper, we propose a method based on GenerativeAdversarial Networks (GANs) to generate fake images that have the same viewpointto solve the different pose problem. Our model first extracts identity-related andpose-unrelated representations from input images and then concatenates the repre-sentation with the pose information to generate the fake image with the assignedpose to deal with the pose variation problem.
|
|---|