Previous weeks - Finnally getting the correct results

Getting the right models

After the productive meeting with Nuria I managed to obtain my own Neural Network models with LSTM1 and LSTM4 for the hole week I have been training models in order to use them in the next steps, use the prediction with the images captures by a camera. For now, I only have been training with modeled samples dataset.

These are the results I obteined:

Testing networks with MODELED samples

I continue the training of different networks with the generated data. Also I add some grapchs to have a visual information about how the test was going on.

Train vs Validation

LSTM 1 with 6000 samples

LSTM 4 with 20000 samples

For the neext week I will be training with raw frames