T. Sim, R. Sukthankar, M. Mullin, S. Baluja. High-Performance Memory-based Face Recognition for Visitor Identification. In Proceedings of the 4tn International Conference on Face and Gesture Recognition, Grenoble, France, March 2000.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....use. The number of features to use in other eigen based methods, such as PCA, is also determined experimentally [55] Since PCA features do not ensure discrimination, removing a few of the dominant PCA features have been noted to yield CHAPTER 2. MRDF FEATURES 65 better classification performance [55,56]; choosing the number of dominant features to discard is another issue in that case. In the MRDF, the only issue is the number of dominant MRDF features to pick, since the discriminatory MRDF features are automatically ordered by their usefulness for discrimination. For the product inspection ....

....case, and with a different training set for each case. Note that in all tests in Table 6.1, we CHAPTER 6. MRDFS FOR POSE INVARIANT FACE RECOGNITION 198 use only one prototype per person. We do not retain all 61 Theta 6 = 366 prototpyes in the training data (as was done in the PCA 2 algorithm in [56]) since this would imply use of six prototypes for each of the training individuals and only one prototype for the test individual; this would bias the modified k NN classifier and P C would decrease. All three modules in our system were used in the final face recognition module P C tests. For a ....

[Article contains additional citation context not shown here]

T. Sim, R. Sukthankar, M. D. Mullin, S. Baluja, and T. Starner, "Highperformance memory-based face recognition for visitor identification," Techni- BIBLIOGRAPHY 217 cal Report JPRC-TR-1999-01, JustSystems Pittsburgh Research Center, vol. 1,

....However, if the detector is applied to a sequence of video frames taken from an interaction with a person, it is quite likely that there will be several correct detections. The face recognition module in TourGuide is applied whenever a face is successfully detected. It is based on the ARENA [23, 22] system developed by Sim et al. from CMU and JustResearch. ARENA uses a nearest neighbor classifier to identify faces by comparing their distance using an L1 norm to labeled example faces (prototypes) taken from a training set. Input faces are correctly identified when they fall within some ....

T. Sim, R. Sukthankar, M. Mullin, and S. Baluja. Highperformance memory-based face recognition for visitor identification. Technical report, Just Research, 1999.

....However, if the detector is applied to a sequence of video frames taken from an interaction with a person, it is quite likely that there will be several correct detections. The face recognition module in TourGuide is applied whenever a face is successfully detected. It is based on the ARENA [11, 10] system developed by Sim et al. from CMU and JustResearch. ARENA uses a nearest neighbor classifier to identify faces by comparing their distance using an L1 norm to labeled example faces (prototypes) taken from a training set. Input faces are correctly identified when they fall within some ....

T. Sim, R. Sukthankar, M. Mullin, and S. Baluja. Highperformance memory-based face recognition for visitor identification. Technical report, Just Research, 1999.

....2.4 Algorithm 3: class constrained test train splits In some applications, the training set is constrained to contain a certain number of items from each class. For instance, the ORL face dataset [8] contains 10 images from each of 40 individuals. The face recognition experiments presented in [6, 9] examine accuracy by varying the size of the training set using 1, 3, or 5 images for each of the 40 classes. In general, we define k to be the number of items required from class k to be in the training set . This section shows how our technique may be applied to this problem, by extending ....

....methods appear frequently (in disguise) in real applications, typically as the final stage of a complex system. Our technique, which is straightforward to implement, is immediately applicable; for instance, Algorithm 3 reduced the time required to evaluate the face recognition systems described in [9] from several hours to a few minutes. This technique may also be used as a fast cross validation component in other machine learning algorithms. We are currently incorporating it into a gradient descent based method for feature weighting with encouraging preliminary results. We plan to extend our ....

T. Sim, R. Sukthankar, M. Mullin, and S. Baluja. High-performance memory-based face recognition for visitor identification. Technical Report JPRC-TR-1999-001-1, Just Research, 1999.

....by one or more agents. A security camera photographs the building entrance every two seconds, and a motion detection algorithm identifies potential scenes containing visitors. Faces from these images are extracted using a neural network based face detector [13] A face recognition system, ARENA [15], examines these face images and attempts to find visually similar matches in its stored database of visitors. Any user interested in receiving notification of visitors runs a user interface agent which is automatically informed when the relevant visitors are identified. This agent also allows ....

T. Sim, R. Sukthankar, M. Mullin, and S. Baluja. High-performance memory-based face recognition for visitor identification, 1999. Submitted for publication. An expanded version is available as Just Research TR-1999-001-1.

....1500 distractor images of faces collected from the web and tagged them with the single label stranger . There has been no noticeable drop in classification performance of known visitors, but unknown visitors are often correctly classified as stranger . For detailed performance statistics, see [24]. 9. Conclusions and Future Work This paper demonstrates that ARENA, a very simple algorithm, can significantly outperform established face recognition algorithms on standard datasets. Unlike the standard PCA based algorithms, ARENA easily handles incremental updates to the face recognition ....

T. Sim, R. Sukthankar, M. Mullin, and S. Baluja. Highperformance memory-based face recognition for visitor identification. Technical Report JPRC-TR-1999-001-1, Just Research, 1999.

....2.4 Algorithm 3: class constrained test train splits In some applications, the training set is constrained to contain a certain number of items from each class. For instance, the ORL face dataset [8] contains 10 images from each of 40 individuals. The face recognition experiments presented in [6, 9] examine accuracy by varying the size of the training set using 1, 3, or 5 images for each of the 40 classes. In general, we define k to be the number of items required from class k to be in the training set 8 . This section shows how our technique may be applied to this problem, by extending ....

....methods appear frequently (in disguise) in real applications, typically as the final stage of a complex system. Our technique, which is straightforward to implement, is immediately applicable; for instance, Algorithm 3 reduced the time required to evaluate the face recognition systems described in [9] from several hours to a few minutes. This technique may also be used as a fast cross validation component in other machine learning algorithms. We are currently incorporating it into a gradient descent based method for feature weighting with encouraging preliminary results. We plan to extend our ....

T. Sim, R. Sukthankar, M. Mullin, and S. Baluja. High-performance memory-based face recognition for visitor identification. Technical Report JPRC-TR-1999-001-1, Just Research, 1999.

No context found.

T. Sim, R. Sukthankar, M. Mullin, S. Baluja. High-Performance Memory-based Face Recognition for Visitor Identification. In Proceedings of the 4tn International Conference on Face and Gesture Recognition, Grenoble, France, March 2000.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC