I want to develop a python script that takes as input a video recording from a dashcam and estimates the speed of the cars in the video.
as I put you will also have the second by second speed, gps coordinates, and gyroscope measurements of the dashcam (mounted inside my car as I'm driving).
Basic method to solving the problem:
- retrain YOLOv3 to also detect licence plates in image. there are many open source licnese plate datasets out there. Yolo is already implelemeted in tensor flow in many open source github repos. you can use any of them.
- draw a bounding bos around each license plate.
- use edge detector to detect 4 edges of license palte.
- you will get real licnese plate dimensions (width and height in cm) as well as focal length of camera as input as well.
- use geometry to figure out coordinates of target license plate in 3d space for each frame. then detect centered of licnese plate.
- you will get frame rate of input video also. you can use this information to calculate speed of licnese plate from frame to frame.
- output in Csv or json format saying coordinates of 4 corners of each license plate, frame number, and estimated speed.
19 freelancers are bidding on average $680 for this job
Hi there, I have read the details I am experienced with Machine Learning, Mathematics, OpenCV, Python. I can help you with this job, Please come to chat so we can discuss this job.
dear Sir, I can do this project. I am a researcher in deep learning, I can assure you that if you work with me once, you will always work with me for these kinds of projects.