Occlusion Detection and Localization from Kinect Depth Images
Bibliographic record
Abstract
Faces captured in a real-world scenario may suffer from large variations in shape and occlusions due to difference in illumination, variation in pose and orientation of a facial image. Automated face recognition or security reinforcement by surveillance techniques would be useless if the faces are occluded. Therefore, face occlusion detection has become very important not only for effective face recognition but also to prevent security threats. In this paper, for the very first time an occlusion detection method is proposed based on the depth information provided by Kinect RGB-D cameras. Uniform Local Binary Pattern (LBP) is used to effectively extract the features from the depth images and SVM binary classifier is then applied to identify the front face and the occluded face. For localizing occluded regions in the face image, a threshold based approach is proposed to identify the areas close to the camera. In the depth images, an object close to the camera has a higher pixel intensity than the object further from the camera. Thus, we assume that occluded regions have lower distance from the camera, i.e. higher intensity values. Based on this hypothesis, we extract the connected component with highest energy values as the potential occluded region from the depth image. The boundary of the detected occluded region is then corrected using the reference front face image. The occlusion detection and localization method have been evaluated on EUROKOM Kinect face database containing different types of occluded and unoccluded faces with neutral expressions. Experimental results show that the proposed method provides an average detection rate of 98.50% for front and occluded face images. We have also compared our proposed method with existing methods that use faces acquired using 3D scanners for occlusion detection.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.001 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.000 |
| Insufficient payload (model declined to judge) | 0.003 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".