YOLO Video Analysis Specialist
Budget: $250 – $750 USD
1- Human detection by YOLO detection model.
How: yolo has an object detection model, ıt can detect 80 different types of objects, including humans. It will give location of humans.
2- (Optional) Selection of a specific human to replace
How: If you want to select a specific human, i will write a code that enables you to select a human from a frame.
3-Tracking the human through the video by YOLO v8 tracking feature.
How: Yolo collaborates "bytetrack" object tracking algorithm. It will track selected humans through all frames.
3- Pose estimation of given Human by YOLO pose. This will give you the skeleton structure of the given body.
How: Yolo can perform pose detection and object detection simultaneously. It will give you the location of various body part of a human. for ex: left shoulder x:16,y15 /right shoulder x:106,y 15.
4- The pose estimation will be matched to the CG character by Blender Python API.
How= Once you get body part location, blender python apı will get those points and map it to the CG character. Now CG characters can perform these movements.
5- Inserting the CG characters through the video and smoothing transition by opencv features such as blurring- HSV matching etc..
How= First we will replace the character with replace anything (this a git hub repo). After that, OpenCV has gaussian blurring for blending the characters' sharp edges.
For light adjustment, the hue of the character will be adjusted according to the general hue of the surrounding environment.
How: yolo has an object detection model, ıt can detect 80 different types of objects, including humans. It will give location of humans.
2- (Optional) Selection of a specific human to replace
How: If you want to select a specific human, i will write a code that enables you to select a human from a frame.
3-Tracking the human through the video by YOLO v8 tracking feature.
How: Yolo collaborates "bytetrack" object tracking algorithm. It will track selected humans through all frames.
3- Pose estimation of given Human by YOLO pose. This will give you the skeleton structure of the given body.
How: Yolo can perform pose detection and object detection simultaneously. It will give you the location of various body part of a human. for ex: left shoulder x:16,y15 /right shoulder x:106,y 15.
4- The pose estimation will be matched to the CG character by Blender Python API.
How= Once you get body part location, blender python apı will get those points and map it to the CG character. Now CG characters can perform these movements.
5- Inserting the CG characters through the video and smoothing transition by opencv features such as blurring- HSV matching etc..
How= First we will replace the character with replace anything (this a git hub repo). After that, OpenCV has gaussian blurring for blending the characters' sharp edges.
For light adjustment, the hue of the character will be adjusted according to the general hue of the surrounding environment.