Purpose of Project

Detection of Pills from images and video. Below you can see the sample detected image.

Labeling Images

Images are labeled by Open Labeling Tool.

Click here for labeled dataset

Splitting Dataset

Here is the python script that splits 90% training and 10% test data. You will have train.txt and test.txt files. Each file has image paths.

Creating and Editing necessary Files

• obj.data (information about number of classes and file paths)
• obj.names (class names)
• yolov3.cfg (editing number of classes and filters)

Since we have 1 class (Pill (turkish meaning: ilac)) our filter must be 18 according to formula.

Creating Google Drive Folder

You need to create a folder shown in that image.

Click here for my Google Drive folder.

References

• How to train YOLOv3 using Darknet on Colab 12GB-RAM GPU notebook and speed up load times
• YOLOv2 ile Kendi Özel Kişi ya da Nesnemizin Algılanmasını Nasıl Sağlarız? — Bölüm 2
• Training YOLOv3 : Deep Learning based Custom Object Detector
• @manivannan_data/how-to-train-yolov3-to-detect-custom-objects-ccbcafeb13d2">How to train YOLOv3 to detect custom objects
• Github Repo