@jeremy I am trying to creat head pose like dataset and train the model. I am finding difficult with two issues.
- When i try to create data bunch my original image is very big(3969,2641) so i try to resize the image to 160,120 by passing size= 160,120 in transform part. But when i checked prediction, the point is predicted out the image frame. My question is when we resize the image while creating data bunch does it also adapt the label coordinate or not…
- I tried by modifing my get_ctr function to return adapted label coordinate but then i am getting another error.
