Multi-neural Networks Object Identification
Autor: | Leonardo M. Bustamante, Iustin Olariu, Jude Hemanth, Teodora Olariu, Daniela López De Luise, Daniel Rivera, Nicolas Park |
---|---|
Rok vydání: | 2020 |
Předmět: |
Artificial neural network
Computer science Real-time computing 0202 electrical engineering electronic engineering information engineering 020206 networking & telecommunications 020201 artificial intelligence & image processing Ultrasonic sensor Image processing 02 engineering and technology Android (operating system) Heuristics |
Zdroj: | Soft Computing Applications ISBN: 9783030521899 |
Popis: | This paper presents an Android prototype called HOLOTECH, a system to help blind people to understand obstacles in the environment. The goal of this paper is the analysis of different techniques and procedures to perform a fast and lightweight model able to detect obstacles in this context. The predictions and analysis are statistically evaluated, and results are used in order to improve the inference results. The model keeps working using low precision images drifted from an Android cell phone supported with ultrasonic sensors. Images are pre-processed on the fly, using multiple neural networks in conjunction with other heuristics, to infer obstacles and their displacement. Results indicate that it is possible to improve the prediction rate and at the same time to reduce extra processing load. |
Databáze: | OpenAIRE |
Externí odkaz: |