When the number of training images decreases (around 50), the results generated by the new perspective are poor. What causes this?
When the number of training images decreases (around 50), the results generated by the new perspective are poor. What causes this?