To train on your computer:

install tensorflow

sudo pip3 install utm python-opencv opencv opencv-python scipy h5py matplotlib python3-tk scipy

sudo apt install python3-tk python-scipy libpcl1.7 build-essential cmake libopencv-dev libpcl-dev libproj-dev

cd rgbd/build

cmake ..

./rgbd_dataset 8 0.60 0.05 60

cd ../..

bash download/download_all.sh

python scripts/gen.py

Check which setting number to use in src/constants.py

src/train.py SETTING_NUMBER NUMBER_OF_STEPS

Train all using

src/train.py all NUMBER_OF_STEPS

The dataset number is 0 for Lip6Indoor and 1 for Lip6Outdoor Again, check which setting number to use in src/constants.py

python scripts/test_lipdataset.py SETTING_NUMBER DATASET_NUMBER

To test all models, run the command:

python scripts/test_lipdataset.py all DATASET_NUMBER

Process SceneNN

Add way to download Streetview

Train all networks and test them

Staging to speed up training

Fix image stat calculations

Use ESC to exit the viewer

Varying the ratio may make it train more quickly. Investigate more. it train more quickly. Investigate more.