Deep Learning based Object Detection

Objectives:

• To detect objects using a pre-trained DNN Mobilenet architecture model

Steps:

1. Download assets and check class labels
2. Model inference
• Load a model and input image
• Detect objects using a forward pass through the network
• Display detected objects with bounding boxes and class labels

Outcomes:

Object Detection using TensorFlow Hub

Objectives:

• To detect objects using a pre-trained object detection EfficientDet D4 architecture model

Steps:

1. Download sample images
• Display sample images
• Define dictionary and map class IDs to class names
2. Model Inference using Tensorflow Hub
• Load a pre-trained model
• Post-process with different detection thresholds
• Display detected objects
3. Formalize the Implementation
• Perform inference on each image and store the results in a list
• Loop over each of the images and display annotated images

Outcomes:

Instance Segmentation with Mask RCNN

Objectives:

• To classify each pixel of a given image to a particular class, and assigning a unique identity to the pixels of individual objects

Steps:

1. Create help functions
2. Download the pre-trained model
3. Download sample images
4. Run inference on the image
5. Visualize the results

Outcomes:

Object Tracking

Objectives:

• To track objects using a tracker class for computer vision

Steps:

1. Download assets
2. Create tracker instance
3. Read input video
4. Setup output video
5. Define bounding box
6. Intilialize tracker
7. Read frame and track object

Outcomes: