1. Experiences of studying in Germany during innovation times
My stay in Ingolstadt for both an exchange and master program during the 2020-2021 brought new and unexpected experiences for both lecturers and students during pandemic times. Conventional education methods used in both theoretical and practical classes had to operate in a completely digital interface in order to ensure safety for all involved.
Deep changes brought by the pandemic provided interesting and creative ways to participate in volunteer and research work at the Technische Hochschule Ingolstadt.
I sought individual accommodation in the old town (city center) of Ingolstadt. While searching for an apartment, I found that most offers that include an area of 25 m² - 30 m² have a monthly cost of 390 € to 550 €. Contracts that have a longer duration tend to be less expensive. However, due to German Laws, it is possible to exit the accommodation if a notice is sent to the landlord three months prior to the exit date – if the contract doesn’t restrict the appliance of these laws. This price includes expenses in water, electricity, heating and internet services. In order to ensure that the target accommodation fills all desired requirements, it is a advised to rent a hotel room for 5 days in order to inspect all the apartment visits scheduled with the landlord via e-mail prior to arrival. After a physical inspection is conducted, it is advised to return to the hotel and read all the contract clauses before signing and sending all the required documents to the landlord.
The city of Ingolstadt requires that after a contract is established for accommodation, the tenant registers in the city hall. The registration process requires a valid passport, a housing confirmation written and signed by the landlord and the signed contract. Documents are sent electronically and the confirmation is sent to the mailbox.
After the process of registration with the town hall, it is recommended to open a student bank account directly with one of the banks. Sparkasse has a zero fee account for students. After scheduling a meeting via e-mail, the client will bring the registration document alongside the university acceptance letter and the passport.
Enrollment at the university has a comprehensive documentation available at the main website. However, equivalence of subjects must be previously defined alongside the student’s home university before each subject is effectively chosen by the student at the start of the semester. After submission of the required documents, each student receives an ID card which grants access to specific laboratories, software services and licenses, the library and printers at the university grounds.
The residence permit requirement can be either requested alongside the THI international office or can be done independently. An e-mail must be sent to the ausländeramt requesting the necessary forms and documents for the specific residence permit. After sending the documents, an appointment will be scheduled to present the documents and collect additional data before emitting the permit.
2. Summary of my research
The purpose of the research stay in THI was to develop an embedded artificial intelligence system using a FPGA (Field Programmable Gate Array) device which can detect and blur car plates and several different kinds of objects such as cars, people, animals, household objects and specific patterns at a much faster rate than a conventional computer equipped with a graphics card. The main advantages of having a device that executes these functions are associated with much lower energy consumption, size reduction for UAV transportation and compatible speeds for applications for AI recognition in high speed cars and aerial vehicles.
Unlike conventional computers, a FPGA device changes the internal integrated circuit logical structure to work simultaneously with multiple functions in parallel, enabling much higher processing times if compared to CPUs or GPUs. The developed work involved programming the internal structure of the FPGA to enable it to run the YOLO v4 algorithm, which is used in state of the art computer vision object detection technologies. Alongside working with conventional the python and C++/C programming languages, in order to configure the device, additional hardware description languages such as VHDL and Verilog were used to build the final program. The YOLO v4 algorithm running on an RTX 2060 GPU reaches a framerate of about 30 FPS, the developed algorithm using the FPGA reaches an average of 320 FPS with the same configuration.
The project will continue after the initial research stay period at THI. Additional data regarding energy consumption and hardware resources usage is still under research and will be necessary to complete my master thesis. Future prospects include integrating a smaller FPGA board in a car in cooperation with the CARISSMA institute in order to make autonomous driving safer and more accurate.
Anderson stayed at THI from October 2020 until February 2021. Supervisors: Prof. Dr. Luis Henrique Assumpção Lolis (UFPR), Prof. Dr. Alessandro Zimmer (THI)