Original Mask-RCNN is mainly used in object detection and scene segmentation.
This project pays attention to another application: try to transfer the original colorful input file (support the input image or video includes at least a person) to the grayscale background while keeps the main character colorful. By using this project, the main character seems Shiny in the window.
Here are some demo images:
|Original Image 1|New Image 1|Original Image 2|New Image 2| | :--: | :--: | :--: | :--: | ||||| |Original Image 3|New Image 3|Original Image 4|New Image 4| |||||
Here are the demo gifs:
|Demo 1|Demo 2| | :--: | :--: | |||
This project needs these packages to support:
Before start, please download the maskrcnncoco.h5 file from the released page and put it in the main folder.
This project provides 2 Ipython notebooks for guidance. The Demo-Image file shows how to process an input image. The Demo-Video file shows how to process an input video. Both files include detailed annotation and explaination. You can execute them step by step easily.
For any question, please use the issue function or connect with [email protected], thanks a lot!