We are dedicated to develop the application of data and AI in various industries
Interface of spatiotemporal data query
Example of the small computer appearance
We are building a spatiotemporal database into a small computer, allowing users to search for data based on spatial and time information.
It does not need to be connected to the Internet, just connect it to the place where you store data such as NAS, it can be implemented on the intranet to ensure that data will not leak out.
Now it is 90% developed, and will be released soon.
GAN Colorization on grayscale images
Historical aerial photographs provide crucial information for efficient long-term environmental monitoring and change detection.
The contribution of color, texture, and lightness features for land monitoring is well known.
However, the early images were mostly grayscale. Therefore, we use Generative Adversarial Network (GAN) technology for automated colorization, then we can get a deep insight from historical aerial photographs.
Here are example coloured image generated by the GAN.
Transforming the 3D Model to HD Map base map
We use Lidar to scan roads, then generate 3D models to draw vector maps such as lane, road edge, whiteline, roadmarks etc. to form an HD Map, which will be used by unmanned systems in the future.
GAN Object Removal
When using 2D images for 3D modeling, the 3D model cannot show the real road conditions because of the cars on the road. In order to solve this problem, in the process of preprocessing, the cars must be removed in advance.
Therefore, we use AI to identify the position of cars and remove them, then use GAN to restore the ground image. Going through this preprocessing before doing the 3D modeling, the results are very good. Here are the examples.
Vehicles can be fully identified, removed and restored ground image even if they are located below the powerline.