Overview / Usage

Project is motivated by one of the pre-Internship project. Definition is that model which can identify or make Bounding box on Custom object . So first tradious task is to prepare dataset with proper Bounding box labels which are compatible with YOLO label file. After that most important step is to select .cfg (configuration file) for training. Then next task is to train from scratch and monitor it over Tensorflow Board. At the end model is able to draw accurate bounding boxes.

Methodology / Approach

Project includes Custom object detection like parrot, monkey etc. which are not present in COCO-80 (dataset of objection detection for different 80 classes). For making dataset and Bounding boxes, YOLO ANNOTATION TOOL is used. Library to be used in the project is 'Darknet' - C/C++ compatible and it's vary fast. Model is trained from the scratch and it uses YOLOv3 configuration file.

Technologies Used

• Darknet

• os, glob

• YOLO ANNOTATION TOOL

​

Repository

https://colab.research.google.com/drive/1Ami91MZrx1cxR_AHoMQmbrNHc_97MAcY

Other links

• Helper Github Repository (https://github.com/AlexeyAB/darknet#how-to-compile-on-windows-using-vcpkg)

• Article - Develop Understanding