Requirement:
Write code that runs on Linux platform (code can be compiled, or integrated or called into other C/C++ program).
Input from following formats (all):
1. H.264 / H.265 native format
2. MP4 format
3. JPEG format
The stream will be multiple FPS ranging form 1 to 30, the code require to process each frame.
Output:
1. Analyze objects on the frame - mark the icon and save into a reference database.
1.1. object type includes human, passenger car, SUV, light truck, truck, bicycle, motorcycle, shopping cart, baby stroller ... etc.
2. Each frame will have output of:
2.1. Object ID, Type identified, type accuracy %, Frame size (640x320 ... etc.), X axis location, Y axis location, X length, Y length, time of data (given), time used to process
Object identification algorithm is based on the reference category similarity which is based on % of the database, which means the larger the database should increase the accuracy and require more time of processing. The code should give an adjustable value of accuracy probability %, and give reference library used.
Object tracking: Code need to reference prior several frames to track the same object and maintain the same object's ID (object enter the stream). Mark the object leaves the stream by missing tracking in the next Y frames (object leaving the stream). The number of continues same object seem in the previous X1 out of X2 frames (can be changed), and mark missing after Y1 out of the next Y2 frames (can be changed).
Please arrange time to interview prior to aware the project. Thank you.
Hello.
I take an interest in your work.
I have full skills in C++ & C, Qt, Android, IPhone, ASP.Net, Web scraping.
Please check my profile and portfolio.
I have to check details of your requirement but I'll try to complete the task as fast as I can.
I'd be happy to keep long partnership,if you'd like.
Looking forward for your kind answer.
Best Regards.
!!!!!!!!!!!!!!! Hello !!!!!!!!!!!!!!!!!
I've a great interest in this project as a Computer Vision expert who have heavily experienced in Image Recognition & Object Classification for over 10 years.
As you can see from my past works & portfolios, I've been specializing in this field and that's why I feel I could be capable of handling this project.
I can fulfill your requirements at high quality level.
Please contact me so that I can discuss with you on more details.
Thanks and Regards,
Richard Minh.
.hello I was building a system similar using Vimeo API and clarifai API . Vimeo for processing the video and separate the frames and clarifai for recognize the elements. this software was built in Google apps script. If you interested in my project I can setting the software for you.