Exploring modality-shared appearance features and modality-invariant relation features for cross-modality person Re-IDentification
نویسندگان
چکیده
• Using the modality-shared and modality-invariant features for cross-modality Re-ID. Designing a new model to extract features. Introducing loss further decrease variations. Most existing person Re-IDentification works rely on discriminative reducing variations intra-modality Despite their preliminary success, such appearance cannot capture enough information due massive discrepancy between RGB IR images. To address this issue, top of features, we relations among different parts (referred as relation features), which help identify persons with similar appearances but body shapes. end, Multi-level Two-streamed Modality-shared Feature Extraction (MTMFE) sub-network is designed, where are first extracted in shared 2D feature space 3D space, respectively. The two then fused into final that both can be reduced. Besides, novel center alignment proposed reduce Experimental results several benchmark datasets demonstrate our method exceeds state-of-the-art algorithms by wide margin.
منابع مشابه
Supplementary Material for “RGB-Infrared Cross-Modality Person Re-Identification”
This supplementary material accompanies the paper “RGB-Infrared Cross-Modality Person Re-Identification”. It includes more details of Section 4, as well as extra evaluations of our proposed deep zero-padding method. 1. Details of Counting Domain-Specific Nodes In the third paragraph of Section 4.2 in the main manuscript, we quantify the number of domain-specific nodes in the trained network in ...
متن کاملSemantic clause types and modality as features for argument analysis1
This work investigates the role of semantic clause types and modality in argumentative texts. We annotate argumentative microtexts with situation entity (SE) classes and additionally label the segments that contain modal verbs with their modal senses. We analyse the correlation both of SE classes and of modal verbs and senses with components of argument structures (such as premises and conclusi...
متن کاملEXPLORING THE TACTILE MODALITY FOR HMDs
Humans commonly are considered to have five separable named senses, called hearing, sight (vision), smell, taste, and touch, providing information about the external world. These senses, their organs, and respective modalities are listed in Table 18-1. The two most dominant senses are sight and hearing. Equipment designers understand human reliance on these two senses and have used them as the ...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
ژورنال
عنوان ژورنال: Pattern Recognition
سال: 2023
ISSN: ['1873-5142', '0031-3203']
DOI: https://doi.org/10.1016/j.patcog.2022.109145