Machine learning approaches:

• Viola–Jones object detection framework based on Haar features
• Scale-invariant feature transform (SIFT)
• Histogram of oriented gradients (HOG) features

Deep learning approaches:

• Region-based Convolutinoal Neural Network (R-CNN) 2014 Ross Girshick, et al, UC Berkeley
• Single Shot MultiBox Detector (SSD)
• You Only Look Once (YOLO) 2015 Joseph Redmond et al
• Single-Shot Refinement Neural Network for Object Detection (RefineDet)
• Retina-Net
• Deformable convolutional networks

Pre-trained models

• COCO SSD MobileNet v1 model

Training from scratch

• collect images
• label images
• run training
• test effectiveness

Transfer Training

• start with a pre-trained model
• add your own images and labels

Python libraries with built-in object detection algorithms

• OpenCV
• TensorFlow

What is a model?

• the output of a training session?
• Data structure of a “model” or “network”?

Inference Graph?

YOLO

Pose Estimation

1. YOLO NAS POS