Face-Landmark-Extraction-Pytorch

Trained with NVIDIA P100 GPU

Data

  • COCO Dataset is used, check thier website for details Link
  • Training and Test data look like this:

Training

  • Transfer Learning is used with Resnet150 for extracting image features(Convolutional Neural Network) and Recurrent Neural Network using LSTM is used.
  • We can see that model is actually learning! :stuck_out_tongue_winking_eye:

Testing

  • Prediction: Looks like my model is working well :grinning:

Conclusion

Overall, with this project, I cannot underestimate the power of deep learning again..

Github