IMG SIZE=48 LR=1e-4 for model-4
Requirements.
1- tensorflow CPU -python2
2- tflearn -python2
3- openCv2 -python2
4- Numpy -python2
5- pandas -python2
6- matplotlib -python2
7- python2
to create model 4 or test on a FER2013 dataset go to data_creation directorydownload FER2013 dataset from kraggle and store in this Directory.
first run the cvsCreate.py then run createData.py
copy train_data.npy and test_data.npy from this current directory to train-test directory
step 1. Go to train-test
step 2. run main.py
step 3. if you are opting for training please follow the above instructions for data creation. Recommended to increase the n_epoc from 3 to 24 in model.py for higher accuracy, staying in train-test dir type the following command on terminal
tensorboard --logdir="log" for graphical view how the model is performing. Type this command once log dir is visble in train-test directory.
step 4. IF YOU WANT TO USE PRE-TRAINED MODEL:
go to train-test directory then into model directory copy all content and save in train-test directory.
step 5. For testing we need a model, first if u want to use the pre-trained model follow step 4 else if you want to create your own model follow step 3. After that select test when you execute main.py.
step 6. If you to run a prediction on live-data :
go to train-test directory but we need a model, first if u want to use the pre-trained model follow step 4 else if you want to create your own model follow step 3. After that select test when you execute main.py.After that select run when you execute main.py.
step 7. If you want to run a analysis on a video file :
go to Video directory, remember to rename the video file into out.mov (right now it only supports mov format).The code is file name sensitive. To run this we need a model, first if u want to use the pre-trained model follow step 4 else if you want to create your own model follow step 3, copy all the new content generated from train-test if u have followed step 3 else simply copy all the contents on model directory to Video directory. Execute start.py. After a while a new file will be generated named out2.mov. In here is your result stored.