How to Improve my recreated yolo algorithm?

Greetings everyone, I have recreated yolo algorithm. I don’t see a lot of documents or tutorials on the internet. So, I study the concept and create my own one.

The model performs just all right. I feel that many things in my model are not correct. It seems I not using the correct loss and my logic on yolo algorithm might be wrong.
Please suggest to me how to improve and make a better model. Sinecerly.

The following files are the code that I recreate
File description

Yolo.ipynb - My recreate yolo algorithum which need to improve
Yolo-DatasetPrepare.ipynb - Prepare dataset for Yolo

Step to recreate

  1. Run all Yolo-DatasetPrepare.ipynb
  2. Run the Yolo.ipynb

Yolo.ipynb (241.1 KB)
Yolo-DatasetPrepare.ipynb (17.8 KB)

I wonder if you have studied the YOLO algorithm introduced at Deep learning specialization? That could be helpful to you.

I haven’t. I just finish on Advanced techniques Specialization and I have take a look at this (What is YOLO algorithum). (What is YOLO algorithm? | Deep Learning Tutorial 31 (Tensorflow, Keras & Python) - YouTube)

Then I combine the knowledge from both sources to create this yolo.

But I am interested in (YOLO algorithm introduced at Deep learning specialization). Can you provide the link to that course? I couldn’t find it on Coursera.


Course 4 week 3 of the DLS specialization, you can audit the videos for free as well. Good luck.

I see. I already finished this course but I will rewatch it. Thank you.

Is there a reason why YOLO is not presented in this specialization?