Welcome to Vattenfall Open data

Explore Vattenfall datasets and innovation projects where the data has been used. Through this portal, Vattenfall aims to promote innovation in data science, and we encourage you to use the data for your own projects and to develop new solutions.

This portal is a pilot scheme for sharing different data and projects. The displayed data and projects may change over time. The data is shared under various licences that permit its use in different applications. You can find more information connected to each dataset.

Fish-AI Open data

Fish-AI is a project to use image recognition to count and analyse fish travelling in fish ladders past hydropower dams. The data in this project is shared under the license CC BY 4.0.

The data contains images (.png) and labels (.txt) containing information about class and bounding box.

To ensure that the Open Data initiative proceeds in the right direction, please take our survey.
To the survey

CC BY 4.0
Sample data (250 MB)
Full package (9 GB)

Project information

It is important to assess biodiversity in rivers in order to understand the environmental impact of generating electricity from hydropower. Hydropower dams are often designed with a fish ladder to bypass the dam, allowing river-travelling fish to jump to the upper part of the river. This is one of the many actions taken to prevent a negative impact on local biodiversity.

Part of understanding the efficiency of the fish ladder and the biological state of the river is to count the fish passing through the fish ladder. This can be done manually or using camera equipment. Historically, infrared cameras have been used for this purpose, but modern developments in AI and camera technology mean that biologists can now use digital cameras to capture additional information more accurately. This information can then be processed and analysed automatically using image recognition software.

Besides reducing the time taken to count fish travelling in the fish ladder, the combined technology also provides more information about the animals. Through our biology experts, we have been able to assess the number of wild versus farmed fish passing the hydropower dam and understand the prevalence of diseases such as fungi that those fish have.

Other research projects have also been conducted in the area, for example, one investigating the impact of imbalances in the dataset on the accuracy of image recognition. Curious readers can find out more here: Convolutional neural network based object detection in a fish ladder: Positional and class imbalance problems using YOLOv3