SiamFC tracker for twin-beam DEP micromanipulation

Demo

From the root of the project, run:

python3 bin/run_video.py \
--video_path data/test.mp4 \
--model_path models/baseline-conv5_e55.pth \
--write_output False \
--visualization True \
--benchmark False

Real-time application

Modify src/config.py with actual settings and make sure 3D_scanning is running (https://github.com/aa4cc/twinbeam-setup/blob/master/JetsonCode/3D_scanning). Finally, run:

python3 bin/run_realtime.py

Experiments

Tracking in green and red channels

Optimal vs direct trajectory

Control algorithms

Reference

[1] Bertinetto, Luca and Valmadre, Jack and Henriques, Joo F and Vedaldi, Andrea and Torr, Philip H S Fully-Convolutional Siamese Networks for Object Tracking In ECCV 2016 workshops

[2] https://github.com/StrangerZhang/SiamFC-PyTorch

[3] https://github.com/bilylee/SiamFC-TensorFlow