GET ALL YOUR ACADEMIC HELP AT ESSAYLINK.NET

Task 

Explore and Prepare datasets

Question 1: visualize all the bounding boxes in the csv file onto the images, an example is like below and answer the question – Are all bounding boxes accurately capture cat faces in all the training images? How do you think the labelling affect your modelling training and performance?

Training

Question 2: Have pre-trained weights been used in your training? Can you compare the training procedures of with and without pre-trained weights, i.e., give the two plots for the training procedures in tensorboard here. What are the difficulties existing in this task?

Do you think PCA can be applied to do data preprocessing? Can you visualize the YOLOv3’s structure, eg how many cnn layers have been used and what are the kernel sizes for different convolutional and pooling layers? The instructions on training your custom YOLO model on the training image dataset can be found in LEO Assessment 2 with file name ‘step-by-step-training-detecting-instructions.pdf’.

Inference
Detect objects in testing image dataset. As there are no ground-truth labelling available to you, what strategies you can use to check your model’s detection before your submission?

GET ALL YOUR ACADEMIC HELP AT ESSAYLINK.NET