anordin95 / object_recognition_bot Goto Github PK

1. Take a whole lotta screenshots. Mac OSX: cmd + shift + 4 + spacebar; then mouseclick
2. Use this tool: https://github.com/tzutalin/labelImg for labelling the images. Follow the setup instructions there.

Once installed:

python labelImg.py ../object_recognition_bot/training_data/

3. Label the data. Helpful hotkeys: https://github.com/tzutalin/labelImg#hotkeys

• w - scan box
• cmd + s - save
• d - next image

Note: use the suggested filename as the save location.

4. Generate the requisite .csv file for training. You will need to modify the variable training_data_dir within the script before running.

python data_input.py

5. Generate class_names.csv. Simply a csv with the label name and a unique id. Here's an example with one label.

$ cat class_names.csv

fishing spot,0

6. Train the net. Example invocation:

keras-retinanet/keras_retinanet/bin/train.py --epochs 20 --steps 100 csv object_recognition_bot/training_data.csv object_recognition_bot/class_names.csv

7. Test the net.

python test_model.py