Tools for data preparations

Prepare environment (linux)

# Run script
bash start.sh

# activate virtual environment
source dataPrep/bin/activate

Prepare environment (windows)

Video to images extraction

python3 video2frame.py -v PATHOFVIDEO -i 5 -c STARTINGIMAGENUMBER -f FRAMERATE

Must have Flags:

• video = video path

Optional Flags:

• frames = frame rate of video. Default value is set to 30.
• interval = Time (in seconds) to capture frame. Default value is set to 1 second
• counter = image number after unique id, default is set to 1.

eg: User want to extract image every 5 second from a 30 fps video on path ../data/dashcam.MP4

python3 video2frame.py -v ../data/dashcam.MP4 -i 5 -f 30 

Video to images extraction using Yolov4-tiny pretrained model

You can use colab notebook here. Please make a copy before you run the codes on colab.

Convert PNG to JPG or JPEG

python3 convert_png2jpg.py --image test.png 

Rename images with random ID

python3 rename.py --dir images

dir = path of directory for captured photos

Data Augmentation

 python3 imgAug.py --dir trial --dest result -R -B --bright 2.0 -F

dir = path of data directory to perform augmentation

dest = path of directory to store result

bright = value of brightness

noise = value of noise

degree = degree of rotation

N = use noise augmentation

R = use rotation augmentation

B = use brighness or darkness augmentation

F = use flip augmentation

Please refer to the code for further explaination

Remove '_jpg', '_JPG', '_jpeg' from filenames

python3 remove_jpg.py --dir PATH_TO_DIRECTORY

dir = path to folder

Train, Test, Split scripts for object detections

python train_test_valid.py --dir experiments --train_out train --test_out test --valid_out valid

dir = directory of database

train_out = path of train dataset output

test_out = path of ts