Apply NN to Video Frames

Overview

Any NN can be integrated into the Video Labeling interface if it has a properly implemented serving app (for example: Serve YOLOv8). The app adds classes and tags to the project automatically.

The application allows you to apply a served neural network to a single frame of the video.

Application key points:

• Connects to Served models in Supervisely
• Supports Object Detection Models
• Supports Instance Segmentation Models
• Supports Semantic Segmentation Models
• Does not support tracking (creates new objects for each frame)

Related Apps

The application supports the following models:

Object Detection:

• Serve YOLOv8
  
• Serve MMDetection 3.0
  

Instance Segmentation:

• Serve Detectron2
  
• Serve YOLOv8
  
• Serve MMDetection 3.0
  

Semantic Segmentation:

• Serve MMSegmentation
  
• Serve UNet
  

How to Run

1. Launch Serve APP, designed for Videos Projects

2. Run Apply NN to Video Frames in the Video Labeling interface:

3. Set the settings and apply NN to the current video frame