This week, we were able to fix most of our problems. The system is now able to detect all chess moves and is ready to be demonstrated. One awkward problem we have is that the chessboard is flipped visually, but all the pieces actually move like they do. Another problem that we mentioned in Slack is that the Raspberry Pi cannot handle the workload of pygame, as shown in our pictures, so we think we cannot use it for our final demo.

We are on schedule with the Gantt chart. We were supposed to finish the integration of UI with the CV this week, and we were able to do it.

Next week, we are going to finish integrating with the FPGA and hopefully only worry about presenting our product.

Below are some successful background subtraction and blob detection that our system can do now.

This week, I was able to fix most of the problems. Now, the board detection works fine. The piece detections also work fine as well, with some constraints. Now that the board is distinctively black and white, the pieces blend too well with the board. As a result, the blob detection is not able to detect the pieces very well. To mitigate this, we have bought pieces that are red and blue so that we can distinguish the pieces from the board. I have tested that this works by coloring checker pieces red. Here is an example of blob detection when a black piece moves from a black square. 

The top case where the knight moves from a black square, it cannot detect the knight properly.

As for the schedule, I am on schedule with the Gantt chart. Now it is time for integration after the interim demo.

Next week, I am going to get started on the Raspberry Pi to see how Jee Woong’s coordination with the UI plays out and just hopefully make the whole process as smooth as possible.

This week I worked on putting the detecting chessboard corners and background subtraction together. Unfortunately, I did not make much progress. It seems like the environment I am in is different from Jee Woong’s which is causing the problem. To mitigate this, I requested a white picture paper and a new black and white chess board, but I am not sure what is happening since I did not get any update from a TA. Here are some of the results I am getting.

The first one is with glare, the second one is without glare, the third one is without glare, but a light from a specific angle. None of them were able to detect the corners. Thankfully, the background subtraction seems to work with or without glare, since they are subtracted anyways.

As for the schedule, I am still on track with the schedule.

Next week, I plan to get this problem fixed and hopefully get a somewhat working system of the chessboard detection.