Skip to content

An E2E tutorial on custom object detection using YOLOv3 with Transfer Learning on Google Colab.

Notifications You must be signed in to change notification settings

NSTiwari/YOLOv3-Custom-Object-Detection

Repository files navigation

YOLOv3 Custom Object Detection with Transfer Learning

Steps:

  1. Prepare your dataset and label them in YOLO format using LabelImg. Once done, zip all the images and their corresponding label files as images.zip.

  2. Create a folder named yolov3 on Google Drive and upload the images.zip file inside it. The directory structure should look something like the following:

yolov3
|__images.zip
   |__ *.jpg (all image files)
   |__ *.txt (all annotation files)
  1. Clone the repository and upload the YOLOv3_Custom_Object_Detection.ipynb notebook on Google Colab.

  2. Run the cells one-by-one by following instructions as stated in the notebook. For detailed explanation, refer the following document.

  3. Once the training is completed, download the following files from the yolov3 folder saved on Google Drive, onto your local machine.

    • yolov3_training_last.weights
    • yolov3_testing.cfg
    • classes.txt
  4. Copy the downloaded files and save them inside the repository you had cloned on your local machine.

  5. Create a new folder named test_images inside the repository and save some test images inside it which you'd like to test the model on.

  6. Open the Object_Detection.py file inside the repository and edit Line 17 by replacing <your_test_image> with the name of the image file you want to the test.

  7. Run the command: python Object_Detection.py.

Output:

GitHub Logo

  • Read the Medium blog for step-by-step implementation.

About

An E2E tutorial on custom object detection using YOLOv3 with Transfer Learning on Google Colab.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published