innovator

Add idea


Calendar

«    July 2019    »
MonTueWedThuFriSatSun
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
 

 

Advert

 

Payment

 

Advert

 

Authorization

Стартап

DateDate: 20-07-2019, 06:25

The technology of classical convolutional neural networks - or as they are also called cCNN - is the main base of algorithms and scripts, thanks to which various neural networks can analyze visual information, correctly process it and fill in the missing fragments. Being a technology that is fairly common at all levels, it nevertheless has a number of significant limitations that do not allow it to be called truly effective and useful in applying it in really complex tasks. For example, in the processing of visual information for artificial intelligence of autonomous cars or forensic neural networks.
It is for this reason that Till Hartman, a talented neuroscience researcher who is a professor of neuroscience at Harvard University, decided to somewhat improve this neural network technology, making it more human-like. In the sense to teach it to think and visualize the space in the way that the human brain does - after all, filling in the missing visual information often takes place under imperfect and often problematic conditions.
He drew attention to how the visual cortex functions, noting its reverse-direct type of visualization and processing of visual information. Joining with the software engineering team, he managed to present so-called recurrent convolutional neural networks that almost completely repeat the process of processing and perceiving the visual information of the human brain, thus filling the empty logical space of the existing display much more efficiently.
Moreover, these CNN reflexes have a truly ambitious level of work, even under the most disturbing circumstances and conditions - for example, the new system of neural networks was tested to determine how well it can complement the blurred image, which it coped well after a few seconds of analysis.