Author: kx1500

What did you personally accomplish this week on the project? 

working on the poster

working on the setup for final demo

improving the recognition system

Is your progress on schedule or behind?

yes we are close to the final demo

What deliverables do you hope to complete in the next week? 

finish our final demo and final report

What did you personally accomplish this week on the project? 

Finished implementing ML algorithm for plate recognition

Setting up environments on AWS server and testing response time

Is your progress on schedule or behind?

My progress is on schedule. Even though our previous OpenCV algorithm works, but we find that ML model has much better accuracy.  I will  finish integrating ML model into our project before the final demo.

What deliverables do you hope to complete in the next week? 

Integrate the ML model part into our project

Work on the transfer of images from pi to AWS server

What did you personally accomplish this week on the project? 

Tried sharpening the images

Tried color recognition algorithm

Researched ML algorithm for plate recognition

Implementing ML algorithm for plate recognition

Is your progress on schedule or behind?

My part is behind schedule because of poor performance of plate recognition algorithm during test situation. As discussed with Professor Kim and TA Manny,  I tried sharpening the image, but it does not work well. So we decide to upload the image to AWS server and run ML algorithm for plate recognition.

What deliverables do you hope to complete in the next week? 

Finish ML algorithm for plate recognition. Set up environment on AWS server for ML algorithm and integrate it to our project.

What did you personally accomplish this week on the project? 

Worked on interim demo video with our group

Worked on improving performance of plate recognition algorithm

Did some research in other recognition such as car model and color.

Is your progress on schedule or behind?

My part is behind schedule because of poor performance of plate recognition algorithm during test situation. As discussed with Professor Kim,  we need to have backup plans when plate recognition does not work.

What deliverables do you hope to complete in the next week? 

Start implementing other backup plans such as car model and color recognition.

What did you personally accomplish this week on the project? 

Tested pi with multiprocessing

Working on pi with multiple cameras

Researching the problem of frame choosing

Is your progress on schedule or behind?

My part is a little bit behind schedule because of the limited numebr of pi we have (need to do multiprocessing using pi) and the frame choosing algorithm (brought up by Professor Kim during Wednesday’s meeting). Professor Kim also mentions a problem: the camera should not keep taking frames while there aren’t any cars coming. I will solve this problem later.

What deliverables do you hope to complete in the next week? 

Integrate the plate recognition part into our project and prepare for our demo.

Keep working on the similarity score algorithm and frame choosing algorithm.

What did you personally accomplish this week on the project? 

I finished the camera and pi setup and worked on the similarity score algorithm.

Is your progress on schedule or behind? If you are behind, what actions will be  taken to catch up to the project schedule?

My part is on schedule. But we are having issue with the http request on pi. Our pi can send http request but cannot receive the content from the request. Roger may need some help in the http request.

What deliverables do you hope to complete in the next week? 

Finish the similarity score algorithm. Help Roger with the http request and prepare for our interim demo.

 

This week I worked on our plate recognition algorithm. I fine-tuned our OpenCV algorithm by cropping the image size and sliding window. However, the plate in PA may contain words other than license plate number. Those words will interfere our result, making our accuracy lower than expected. Next week I will continue to solve this problem. Hopefully, we are able to solve this probelm. Otherwise, we have to upload the images to AWS and do the plate recognition using ML-based algorithms.

This week I worked with my team to visit the East Campus Garage to determine the hardware set up for our MVP presentation. Also, I proposed some abnormal cases that we need to handle during our discussion with Professor Kim. For the CV algorithm part, I have tested openCV algorithm for car plate recognition. However, the recognition accuracy is lower than we expected. So, next week I’m going to test some ML-based recogtion methods to see if our requirements can be reached (processing time will be a major consideration). If ML-based methods can fulfill our requirement then I’m on schedule. If none of the methods work, then I will need to speedup and consider other solutions.