AI Hub Workshop 30.8.2019: Object Detection
The introductory examples given by Associate Prof. Heikki Huttunen were about people detection and machine vision. As a first example, Tampere University has placed a Raspberry Pi based people detector at Vapriikki Museum area in Tampere. This technology is used to measure eg the frequency of visits. The edge device does not identify individuals or store their identities. The detector is designed with privacy as a main driver.
Another example given on applied object detection is provided by Visy Oy that uses machine vision to text and number recognition. The company has developed a system that reads the number plates that identify eg train cars or cargo containers at port areas.
Thirdly, Tampere University campus has a number of cameras and detectors that scan the area using bounding box detection and semantic segmentation. Both of these can be overlapped in order to get a more extensive and accurate data representation of the environment.
Huttunen also introduced a person detection Python demo on Jupyter Notebook/Spyder, using Open CV. Towards the end of the workshop, participants had their own people detectors running on their devices.
- 9:00 Welcome coffee
- 9:30 Fundamentals of object detection in the machine learning paradigm.
- 11:00 Hands-on session 1: Classical object detection using scikit-learn and opencv libraries.
- 11:30 Lunch (on your own)
- 12:30 Deep learning for classification
- 14:00 Deep learning for object detection
- 15:30 Hands-on session 2: Using tensorflow object detection API.
- 16:00 Closing