Excerpts of video segmentation:
Due to github limits you can find weights, videos and output of video.py here.
For colab realization:
- cuda_11.1
- python version- 3.7.13
For another one(local training):
- tensorflow-2.8.0
- Python 3.7-3.10
- cuDNN 8.1
- CUDA11.2
References
1. https://www.kaggle.com/datasets/tapakah68/segmentation-full-body-tiktok-dancing-dataset - TikTok full body segmentation dataset;-
https://www.youtube.com/watch?v=J_XSd_u_Yew&t=126s&ab_channel=DigitalSreeni - Semantic Segmentation;
-
https://arxiv.org/abs/1505.04597 - U-Net;
-
https://github.com/maxvfischer/keras-image-segmentation-loss-functions - Types of loss function for semantic segmentation;
-
https://pycad.co/the-difference-between-dice-and-dice-loss/ - Dice metric ans dice loss;
-
https://www.youtube.com/watch?v=qrL22HEaUGA&t=3840s&ab_channel=IdiotDeveloper - UNet Semantic Segmentation;
For solving GPU run out:
-
https://www.youtube.com/watch?v=hHWkvEcDBO0&t=196s&ab_channel=AladdinPersson;
-
https://stackoverflow.com/questions/36927607/how-can-i-solve-ran-out-of-gpu-memory-in-tensorflow;
-
https://www.tensorflow.org/guide/gpu#limiting_gpu_memory_growth;
-
https://www.tensorflow.org/api_docs/python/tf/config/experimental/set_memory_growth;
-
https://stackoverflow.com/questions/65385221/running-out-of-memory-so-quickly-tensorflow-gpu.
Evelina Alexiutenko - LinkedIn - evelina.alexiutenko@stud.onu.edu.ua