Current location - Education and Training Encyclopedia - Educational Knowledge - Which software is better for machine vision?
Which software is better for machine vision?
1. The traditional machine vision software, represented by Halcon and VisionPro, customizes various parameters according to the target characteristics by calling various algorithm factors. It is good at solving quantitative problems mainly based on measurement and judging whether there is a simple problem, but for the needs of complex detection, the missed detection rate/false detection rate is high. Although the price of software has been falling again and again, the price of domestic software is only a few thousand yuan or even free, but the cost of early development and late maintenance is very high, so it faces great limitations in use.

2. The deep learning software represented by VIDI and Sualab of cognex (acquired by cognex) encapsulates the marking, training and testing process of deep learning in the form of PC software, which lowers the threshold for users to use deep learning technology. Mainly solve the complex defect detection problem that traditional vision can't solve, or use deep learning annotation method to improve development efficiency. However, from the practical application effect, the current deep learning mainly focuses on supervised learning, which requires a lot of work such as model selection and parameter adjustment for different scenarios, and cannot be universally solved for all scenarios. Therefore, it does not solve the problems that traditional algorithms need to be customized and developed, users have a high threshold for use, and ordinary engineers generally have no basis for parameter adjustment of deep learning.

3. Beijing Moment Vision Intelligent Technology Co., Ltd., founded on 20 17, has received nearly 10 million yuan of financing, innovatively provided cloud-based deep learning tools for industrial vision, and realized the process of online marking, training and testing. Relying on a large number of deep learning model libraries of different scenes accumulated in the cloud, after users upload and mark pictures, the cloud automatically matches the optimal deep learning model and parameters for training, and users do not need any operation to test online. At the same time, in business, Moment Smart will directly free cloud tools. After downloading the sdk, you only need to buy a Lisence dongle to run locally, and you are committed to applying deep learning technology to all industrial visual scenes.