yabufarha / anticipating-activities

Other
77 stars 19 forks source link

Running without GT observations #4

Closed ClaraBing closed 5 years ago

ClaraBing commented 5 years ago

Hi Yazan,

Thank you so much for sharing the code! I used the model on our dataset with GT observations and it works great!

May I ask if there are also code for running the model without GT observations? Thank you very much!

yabufarha commented 5 years ago

Hi,

Happy that you found the code helpful. For running the code without GT observations, we used the model from https://github.com/alexanderrichard/squirrel to get the predicted action labels of the observed part and then run our trained model with these labels. The code is in examples/cvpr2017 Note that this model is for weakly supervised case but we train it fully supervised. Related publications: Richard et al. Weakly Supervised Action Learning with RNN based Fine-to-coarse Modeling. CVPR 2017 Kuehne et al. A Hybrid RNN-HMM Approach for Weakly Supervised Temporal Action Segmentation. PAMI 2019