Model Returning Poor Results: Course 1 Week 1

I’ve built my model as best to the instructions as I can. The loss for both training and validation is below 1, and is gradually reduced. However, my iou is below 0.5 for overwhelming majority of results.

After evaluation,
Loss: 0.08514761179685593
Number of predictions where iou > threshold(0.5): 1
Number of predictions where iou < threshold(0.5): 499

1 Like

Here are my results visualized. Model is not learning

Which assignment is this? I dont see any IOU function there in C1 W1.

1 Like

Hello danieldacruz7!

You have posted on the forum of the TensorFlow: Advanced Techniques Specialization.

Course 1 of this specialization is Custom Models, Layers, and Loss Functions with TensorFlow.

And throughout this specialization I don’t recall any assignment involving the object detection problem appearing on the image you provided. It’s possible you posted your question on the wrong forum.

If you post more details about the code you’re running I would be glad to help, as it is a topic I’m well familiar with, however it would be much more appropriate to post the whole thing on the forum related to the correct course. There you will be assisted by mentors who have ready access to those assignments and will be more apt to help!

1 Like

@danieldacruz7, I moved this post to Course 3, Week 1, since Advanced Computer Vision with TensorFlow is Course 3 in the specialization, and it will be easier for people to find here. In the future, if you’re not sure of the course number, one way to check is to look at the title of the lab file. It should start with Cn where the n is the course number - in this case, C3W1_Assignment.ipynb.

As far as the problem you’re seeing - one thing I notice in the visualization you shared: all of your prediction boxes are in the upper left corner. That looks very suspicious. I’d suggest starting with printing out the (xmin, ymin, xmax, ymax) for predicted_bboxes and then work backwards from there to try to track down what’s going wrong.

2 Likes