Computer Vision
12 important questions on Computer Vision
What are problems for Object recognition?
Different in color, viewpoint, lighting, surround, occlusion
-Object differentiation: We recognize objects from thousands of similar objects
What are invariant features we need for recognition?
- Y-Junctions: 3 surfaces meet
- Parallelism, Symmetry
-Geons
What role do curves play in object recognition?
- Higher grades + faster learning
- Never study anything twice
- 100% sure, 100% understanding
What role do retinal ganglion cells play in object recognition?
What are view-invariant characteristics?
- Y Junctions
- T Juncitions
What do T Junctions mean?
How do we learn our typical invariants?
Backpropagation Neural Network:
Give the system something to help it classify (E.G. Habitat = water, higher chance its a fish).
If you do this with alot of animals, the system will generalize: recognize animals that the network was not trained on.
Convolutional Neural Networks:
-
How do CNN's learn their data?
- On the basis of very large image set
In CNN what is the hierarchy?
- Higher level features: Differ depending on what you want it to see
What are the most important improvements in CNN?
2. No a-priori filter sets (learning from inputs)
The question on the page originate from the summary of the following study material:
- A unique study and practice tool
- Never study anything twice again
- Get the grades you hope for
- 100% sure, 100% understanding