Skip to content

Latest commit

 

History

History
22 lines (17 loc) · 1.05 KB

README.md

File metadata and controls

22 lines (17 loc) · 1.05 KB

Activity Recognition from Egocentric Photo-Streams

Project to recognize human activities from egocentric perspective.
Pre-trained models are exploited for the recognition.

Check README.md located in archive for more details.

Pre-trained Models

Download the folder with the pre-trained models from this link.

Requirements

Please refer to requirements.txt.

Notebook

25122021.ipynb

  • Cell splits the input video located in ./input/ to frames stored ./input/frame/.
  • Cell read frame by frame located in ./input/frame/, predict labels, put predicted labels in the dedicated frame as text.
  • Cell read the recognized frames located in ./input/frame/ and build video stored in ./output.

Issues

Unable to find the correct tensorflow-gpu and keras versions.
Some pre-trained models are not well trained.
Maybe retrain the models.