YOLOv3 Pytorch Implementation

Brief Description

This is an implementation of the object detection architecture "YOLOv3" using the deep learning framework "Pytorch". YOLOv3 is a neural network architecture for performing object detection on images and videos, it's a refinement of the previous YOLO9000.

Utilized Technologies and Frameworks

Pytorch.
Basic Python Data Manipulation Packages (Matplotlib,....etc).

Repository Structure

1) Main Files:

yolov3_model.py: contains the definition of YOLOv3 architecture.
yolov3_train.py: contains the training setup of the network.
yolov3_validate.py: contains the validation setup to test your trained model.
yolov3_detect.py: contains the setup to run inference on images using the trained model.

2) Configs:

yolov3.cfg: contains the configuration of the neural network architecture along with hyperparameters.
yolov3.data: contains the paths to the data and number of classes.
yolov3.names: contains the names of the classes.

3) Data Preparation:

img_prep.py: used to prepare the dataset images for the model.
label_prep.py: used to prepare the labels (bounding boxes) for the model.

4) Utilities:

data_classes.py: contains the classes for the dataloader.
helper_func.py: contains some functions to parse config files.
yolov3_utils.py: contains some utility functions: non-max suppression, output preparation and some metrics.

How to Use it

Run data preparation files providing train path, labels path and test path.
Edit config files at the mentioned lines beginning with (//--------->).
Optionally, Edit some hyperparameters in yolov3.cfg.
Run yolov3_train.py for training.
Run yolov3_validate.py for validation.
Run yolov3_detect.py for Inference.

Acknowledgement

"YOLOv3: An Incremental Improvement" Research Paper: https://arxiv.org/abs/1804.02767
I have learned a lot and utilized a lot of resources from this repository: https://github.com/ayooshkathuria/YOLO_v3_tutorial_from_scratch

Thanks a lot ^_^