2019 IEEE International Workshop on Information Forensics and Security (WIFS) 2019
DOI: 10.1109/wifs47025.2019.9035102
|View full text |Cite
|
Sign up to set email alerts
|

Image Semantic Representation for Event Understanding

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
1
1
1

Citation Types

0
2
0
2

Year Published

2020
2020
2024
2024

Publication Types

Select...
4
2

Relationship

3
3

Authors

Journals

citations
Cited by 7 publications
(4 citation statements)
references
References 12 publications
0
2
0
2
Order By: Relevance
“…Figures S7 and S8 show the obtained results when training the networks with small quantities of original representative images, using cross-entropy and contrastive, respectively. In this case, we compare N (512, 128) and N (1024, 512) to the semi-supervised method ESS [1]. Even with reduced quantities of training images, the components combination N (512, 128) and N (1024, 512) produced the best MAP results, resulting in an improvement of more than 10 percentage points in the smallest set (with only 10 representative images).…”
Section: Experiments Ii: Varying Training Quantitiesmentioning
confidence: 99%
“…Figures S7 and S8 show the obtained results when training the networks with small quantities of original representative images, using cross-entropy and contrastive, respectively. In this case, we compare N (512, 128) and N (1024, 512) to the semi-supervised method ESS [1]. Even with reduced quantities of training images, the components combination N (512, 128) and N (1024, 512) produced the best MAP results, resulting in an improvement of more than 10 percentage points in the smallest set (with only 10 representative images).…”
Section: Experiments Ii: Varying Training Quantitiesmentioning
confidence: 99%
“…Já as CNN podem ser aplicadas de variadas formas. Dois exemplos encontrados na literatura são: combinação de diferentes redes, sendo cada uma especializada em identifi car uma característica diferente (como pessoas, lugares, ou objetos), e a fi ltragem baseada na combinação desses aspectos (Rodrigues et al, 2019); e fi ltragem em três passos: uma rede fi ltra imagens que são inerentemente irrelevantes (e.g., desenhos), depois imagens duplicadas são removidas por Perceptual Hashing e, por fi m, as imagens restantes tem sua relevância avaliada por outra rede (Nguyen et al, 2017).…”
Section: Coleta E Filtragem De Dadosunclassified
“…carried out using two pressure cooking bombs that caused explosions near the finishing line of the race. The attack killed three persons and injured several hundred others[22]. fire in a densely populated district in Dhaka, Bangladesh.…”
mentioning
confidence: 99%