Hello everyone,
I found a dataset on Kaggle and I wish to apply image regression on it. But I’m not sure how to go about it. The dataset has 2 types of files. Images of cats stored as .jpgs and another .jpg.cat file corresponding to that image. The cat file has coordinates for the ears, nose, and eyes of the cat separated by spaces. How can I use image regression on this dataset. I don’t quite understand the code from lesson 3. Any type of guidance would be really helpful.
Thanks and regards.
Which part(s) of lesson 3 are you struggling with? It’s going to be difficult for people here to help, without knowing more details/descriptions of what you have difficulties with…
The dataset you described actually sounds very similar to the dataset used in lesson 3 (regression of point coordinate on face). I think it would be reasonably straightforward to modify the lesson 3 Jupyter notebook to work with your dataset. Perhaps you can start with just one thing from your dataset (maybe the nose coord?) so that it’s completely similar to the lesson 3 data. When you get that working, I’d imagine you’ll be familiar enough with the notebook and workflow to expand to all the points in your dataset.
Any question/errors you have, please ask here, with relevant details as described in this thread.
Hey, thank you for the response and sorry for the vague question. I looked at the lesson 3 notebook again and realized that there was a lot of pre processing being done to get the co-ordinates of the point which I don’t have to do. However, I’m not able to plot multiple points on the same image. Can you tell e how can I pass a list of tuples or something similar to plot all the points and then train a model for that?
I’m able to do it for a single point but not more points.
Thanks.
Hey, I did not. However I found that the co-ordinates were (x,y) values which was causing an issue. Fastai uses (y,x). A simple flip of the co-ordinates should make it work.
Cheers