Challenges in visual object recognition
Matching and learning visual objects is challenging because of the following factors.
- Illumination conditions
- Object pose
- Camera viewpoint
- Partial occlusions
- Unrelated background clutter
- Computational complexity
- Scalability
State of the art of visual object recognition methods
A good feature extraction method needs to have the following properties.
- Being able to extract local invariant features (e.g. scale invariance, affine invariance)
- Fast to extract
- Robust to viewpoint variations
- Retaining enough discriminative power to allow for reliable matching
References
Kristen Grauman and Bastian Leibe (2011) . Visual Object Recognition (Synthesis Lectures on Artificial Intelligence & Machine Learning). Morgan & Claypool.
No comments:
Post a Comment