Let's now test our dataset loader. I assume that you have already finished step 1 (rescaling the videos).
You can now run the following line:
cd loader
python test.py --dataset vlog
# Generic command
python test.py \
--dataset <DATASET-NAME> \ # vlog or epic
--root <ROOT-DIR-DATASET> \ # on the above example it is ../data/vlog
--t <VIDEO-LENGTH>
If you have a new PNG file called img.png
in this directory it means that you are able to iterate over your videos!
The first time you run the test.py
file it should take some time because it is looping over the directory to make sure that all the videos are present and retrieve their label and length.
ps: the code for visualization of the masks has been adapted from detectorch
You can start training and evaluating on this dataset => README_training