Watched OpenCV tutorials for beginners https://www.youtube.com/watch?v=kdLM6AOd2vc&list=PLS1QulWo1RIa7D1O6skqDQ-JZ1GGHKK-K 🤗What is YOLO? You Only Look Once is an object detection algorithm in which only a single convolutional network predicts the bounding boxes and their corresponding class probabilities. Blogs: https://medium.com/analytics-vidhya/everything-you-need-to-know-to-train-your-custom-object-detector-model-using-yolov3-1bf0640b0905 https://medium.com/@akarshzingade/logo-detection-using-yolov2-8cda5a68740e
Read this blog and implemented the same on my system. Link: https://towardsdatascience.com/yolo-object-detection-with-opencv-and-python-21e50ac599e9 My implemention provided above.
✨Implemented a small object detection model on coursera Course link: https://www.coursera.org/projects/real-time-object-detection-yolo Certificate: https://coursera.org/share/fa3ded4fcd97b5032653d2dd77bee9ee
Gone through this blog for video processing https://towardsdatascience.com/yolov3-pytorch-on-google-colab-c4a79eeecdea
Steps required to implement object-detection model:
- Collecting the dataset
- Pre-processing and Augmentation
- Annotation
- Training our model
- Obtain the weights
Started implementing the above in roboflow https://blog.roboflow.com/training-a-yolov3-object-detection-model-with-a-custom-dataset/
For Preprocessing and Augmentation: https://blog.roboflow.com/why-preprocess-augment/
Augmentation basically creates new training images from the existing images. This task is done using roboflow itself.
Studied about overfitting and underfitting: https://www.kaggle.com/dansbecker/underfitting-and-overfitting
Read about batch and epoch which i came across while training in colab. https://machinelearningmastery.com/difference-between-a-batch-and-an-epoch/
Images are annotated in roboflow and then trained using google colab . Colab link : https://colab.research.google.com/drive/1D4X949Ol3LDQtj2mfZPuKesR4k1fjZho
This model doesn't works fine so first tried to implement on just two classes in colab. Colab link : https://colab.research.google.com/drive/1Kc7u6SNvIc5D4ZyjjK_d5IMSGe3a-3yA