yaorong0921
combined inside and outside prediction
Hello. Is the script for training using outside and inside videos not available?
Hi hivabhagat, to combine the features from inside and outside videos is actually straight-forward. I first saved the features from two video streams beforehand to save some computation time, and then used several layers to process to get the final results. I can try to find the script for this. It has been a while and I might need some time.
If you need help with the model itself, pls let me know.
If you could find the script that would be great. I am trying out an approach for my thesis and i want to use your proposed model as a baseline
could you provide codes for getting the accuracy and outside video or code to run the model on a demo video? It would really help me out. I would really appreciate it.
could you provide codes for getting the accuracy and outside video or code to run the model on a demo video? It would really help me out. I would really appreciate it.
Hi there, could you please check the script main_outside.py to run the code on outside video? For the demo, I run the script to get the results for test samples at each timestamp, and then just add the predicted action to the corresponding video. Hope it helps you.
I understand that we get the predicted frame from the convLSTM. Is there a way to calculate accuracy? You mention accuracy and F1 scores for outside video but I cannot seem to generate them using the function in utils. I am interested in reproducing the results of your models and it will really help me out if I can get all the codes. Could you email them to [email protected]?
Hello, is it correct that main_inside.py and main_outside.py output the features extracted from inside and outside respectively? Then they are input into a training model to output intention. My confusion is which script is the final training model that can output intentions?
Since I am working on a project taking the predicted driver's intention as input for other training aims, I wonder which model of yours can be used as input directly. Thanks a lot.
Hi @yaorong0921 thank you for sharing the one drive link for weights, I see that the code for running the model using both inside and outside features is missing, please would it be possible to share the complete to [email protected]? Also how to generate results like the ones you have in the demo folder from the one drive link?
@Hirva-Bhagat were u able to get the code for this work, if so please would you share them?