Video Analytics project
Budget: $250 – $750 USD
Video Analytics Project
Project Description
The project is to provide video analytics on any video stream. The video analytics will identify any person, a person in motion, a object (car or vehicle or other objects) in motion or a animal or any moving object in the video frames. (this is called “R1 frame”)
from the rest of the frame (also the rest of the frame is to be called “R2 frame”). For each “R2 frame” from each video frame, continue to a diff of “R2 frame” to the next “R2 frame” and provide if there is a difference or not. If there is no difference, you will keep the “R2 frame” and mark it with a name tag “Nodiff”. If there is difference you will mark the difference and mark After identifying this, you will extract out this type of moving objects , animals or person and separate it it with a tag “DiffTrue”. So each frame will have Nodiff1 , Nodiff2 .. or Difftrue1 , Difftrue2 ….so on.
We also want to make sure that any video stream can be initially classified by a user with a 30 sec video to say which is” relevant” information and which is not. You will use this information to do the object/animal/person detection.
Once this extraction is done, when a user wants to replay, we should be able to combine the “R1 frame” with the “R2 frame” to show the video as in original stream.
Also it has to be able to take multiple video streams for processing. It is important to make sure, if it can be achieved through cpu and try not to use GPU. So code has to be highly efficient and scalable
If you have the necessary skills and experience in video analytics and pattern recognition, please submit your proposal.
Project Description
The project is to provide video analytics on any video stream. The video analytics will identify any person, a person in motion, a object (car or vehicle or other objects) in motion or a animal or any moving object in the video frames. (this is called “R1 frame”)
from the rest of the frame (also the rest of the frame is to be called “R2 frame”). For each “R2 frame” from each video frame, continue to a diff of “R2 frame” to the next “R2 frame” and provide if there is a difference or not. If there is no difference, you will keep the “R2 frame” and mark it with a name tag “Nodiff”. If there is difference you will mark the difference and mark After identifying this, you will extract out this type of moving objects , animals or person and separate it it with a tag “DiffTrue”. So each frame will have Nodiff1 , Nodiff2 .. or Difftrue1 , Difftrue2 ….so on.
We also want to make sure that any video stream can be initially classified by a user with a 30 sec video to say which is” relevant” information and which is not. You will use this information to do the object/animal/person detection.
Once this extraction is done, when a user wants to replay, we should be able to combine the “R1 frame” with the “R2 frame” to show the video as in original stream.
Also it has to be able to take multiple video streams for processing. It is important to make sure, if it can be achieved through cpu and try not to use GPU. So code has to be highly efficient and scalable
If you have the necessary skills and experience in video analytics and pattern recognition, please submit your proposal.