YOLOv4-based Object Detection Model.
The model was trained with 5083 images in total. On that 4499 images from day-time IR images and 584 images from night-time IR images. The model can be on tested 20% of the day-time image dataset(1124 images) and 8000 images night-time IR data set.
The model trained to detect classes as,
"category_id": 0 for "people"
"category_id": 1 for "buggy"
"category_id": 2 for "motorcycle"
"category_id": 3 for "car"
"category_id": 4 for "ATV"
"category_id": 5 for "bus"
"category_id": 6 for "truck"
"category_id": 7 for "van"
The included files are:
-
Weights
The file contains the model.
-
Testing day IR images
The images used to test the model.
-
Testing night IR images
The images used to test the model.
-
Python scripts
(i) detector_output.py
This file is used to output the detected objects in a JSON file format for day IR image
(ii) detector_output_1.py
This file is used to output the detected objects in a JSON file format for night IR image
The model was tested on a Pycharm IDE.
The packages used to run the Python script.
(1) opencv-python https://github.com/opencv/opencv-python
(2) NumPy https://github.com/numpy/numpy
(3) glob
(4) random
(5) JSON