Multi Object Tracking

Multi-Object Detector and Tracker

Object Detection

Deep learning based Real-time Object Detection

Object Tracking

Deep learning based tracker trough Multi-camera View

Object Detection

  • A deep learning-based approach

  • Based on the state-of-art detectors

  • Robust (trained on the dataset with noise, masking, corruption)

  • Acceptable Accurate

  • Support different type of Objects

  • Tested on different Object such as Pedestrians and Vehicles

  • Our detector needs optimization for processing time

Object Classifier

  • Based on SimpleNet and Efficient Net

  • Near Real-time

  • Accurate enough (about 95%)

  • Robust against camera changes, light conditions, weather conditions,etc

Multi-Object Tracker

    • Combining Tracking and Detection

    • Multi-object Tracker

    • Online learning (learning based)

    • Real-time tracker

    • CNN-based feature extractor and using Triplet-loss

    • Using Kalman filter for Motion Estimation

Scalability and Quality Assurance

  • The solution is highly scalable and can be used in several applications

  • It does not depend on the camera type, position or situation

  • The same solution can be applied in the similar scenarios