A real-time interactive restoration system for intraoral digital videos using Segment Anything Model

Code Vident-lab Dataset

Workflow of the proposed interactive system.

Comparison between input videos and restored videos.

Video Super-Resolution

Using BasicVSR (from 256*256 to 480*480).

Video Super-Resolution

Using BasicVSR++ (from 256*256 to 720*720).

Video Interpolation

Using TOFlow (from 30 fps to 45 fps).

Video Interpolation

Using FLAVR (from 30 fps to 60 fps).

SAM Results

Segmentation results.

Class Activation Mapping Visualization

Using Grad-CAM to visualize the activation region.

Related Links

There's a lot of excellent related work: Segment Anything (SAM), MMagic, BasicVSR, BasicVSR++, TOFlow, FLAVR, Image Colorization, HuggingGPT, Segment Anything in Medical Images, How Segment Anything Model (SAM) Boost Medical Image Segmentation?, Validation of assessment of intraoral digital photography