Evaluating AAM fitting methods for facial expression recognition
read more
Citations
Affect Detection: An Interdisciplinary Review of Models, Methods, and Their Applications
Robust Discriminative Response Map Fitting with Constrained Local Models
Automatic Facial Expression Recognition Using Features of Salient Facial Patches
Static facial expression analysis in tough conditions: Data, evaluation protocol and benchmark
Automatic Facial Expression Recognition Using Features of Salient Facial Patches
References
A Practical Guide to Support Vector Classication
Active appearance models
Affective Computing
Active Appearance Models
Related Papers (5)
Dynamic Texture Recognition Using Local Binary Patterns with an Application to Facial Expressions
Frequently Asked Questions (18)
Q2. What are the contributions in "Evaluating aam fitting methods for facial expression recognition" ?
This paper evaluates various Active Appearance Model ( AAM ) fitting methods, including both the original formulation as well as several state-of-the-art methods, for the task of automatic facial expression recognition.
Q3. What have the authors stated for future works in "Evaluating aam fitting methods for facial expression recognition" ?
A future direction for this research is to take advantage of accurate ID fitting algorithms to tackle the problem of pose-invariant expression recognition.
Q4. What is the future direction for this research?
A future direction for this research is to take advantage of accurate ID fitting algorithms to tackle the problem of pose-invariant expression recognition.
Q5. How many real-time AAM trackers were used for PIFER?
For PDFER, 30 real-time AAM trackers, one for each subject, were trained separately, whereas, for PIFER, a single real-time AAM tracker was used that was trained to track the facial features across the 30 speakers in the database.
Q6. What are the normalisation distances used to normalise the feature vector?
Vη and Hη are the normalisation distances used to normalise the feature vector with respect to the varyingsize of faces of different people.
Q7. What is the way to solve the problem of large variation in shape and texture?
if the object exhibits large variation in shape and texture, its performance deteriorates because of the assumption of a fixed linear update model which can be too restrictive.
Q8. What is the way to build a linear update model?
since the work presented in this paper deals with the problem of AAM fitting, building the linear update model based on the mean appearance parameters, which on average are closer to the true parameter values, is an optimal choice.
Q9. What is the definition of a fitting algorithm?
The AAM fitting algorithms can be broadly classified into two categories [21]:Generative fitting deals with the problem of fitting as minimisation/maximisation of some measure of fitness between the model’s texture and the warped image region.
Q10. under what conditions does IEBM outperform HFBID?
Under extreme conditions, IEBM marginally outperforms HFBID by taking advantage of its linear regressor, whose predictive domain is much simpler than that of the nonlinear regressor used by HFBID.
Q11. Does the use of different kernels affect the performance evaluation for various AAM fitting methods?
Since the authors treat PDFER and PIFER as separate problems, the use of different kernels does not affect the performance evaluation for various AAM fitting methods.
Q12. How does IEBM perform in comparison to PIFER?
In comparison for PIFER, on increasing the perturbation to ±25 pixels for initialisation, FJ, SIC, and RIC are unable to converge, however, both HFBID and IEBM still maintain high accuracy.
Q13. How is the accuracy of the perturbation compared to the other algorithms?
As the perturbation is increased to ±10 pixels for initialisation, a slight dip in accuracy for FJ, POIC and SIC is observed, however, HFBID, IEBM and RIC are able to maintain almost the same accuracy for PDFER, whereas for PIFER, the accuracy of FJ, SIC and RIC deteriorates even further.
Q14. What is the disadvantage of the FACS coding scheme?
one of the inherent difficulties with the FACS coding scheme is that it requires a highly trained human expert to manually score each frame of a video.
Q15. How accurate was the scoring for gender based expressions?
The experiments were performed on still images and a maximum accuracy of 79.9% for gender based expression classification and 76.4% for gender independent expression classification was reported.
Q16. What is the inverse compositional method for AAM fitting?
In [1], the idea of the inverse compositional method for AAM fitting is extended further by using an M-estimator (robust penaliser) instead of the least squares fitting criterion, resulting in an iteratively reweighed least squares fitting scheme.
Q17. How accurate is the initialisation of PDFER?
Going a step further for PDFER, on increasing the perturbation to ±30 pixels for initialisation, FJ, POIC, SIC and RIC are unable to converge, whereas, HFBID and IEBM still maintain high accuracy with IEBM performing better at 94.19% accuracy compared to the HFBID’s which achieves 88.03% accuracy.
Q18. What is the process of recognising facial expressions?
After the system has been trained for recognising a set of expressions, it accepts an image as input, followed by the same process of tracking facial features and extracting a feature vector.