Batch Image Processing in Facial Detection Applications
Bibliographic record
Abstract
Batch image processing for facial detection involves running a facial detection algorithm, in parallel, on batches containing multiple images rather than serially on a sequence consisting of single images. Batch image processing is crucial in live-video facial detection applications where the real-time processing of many frames is required. The performance of a facial detection application can be drastically improved when facial detection is done in parallel on batches containing multiple images. In this work, we analyze the performance gain due to running a GPU-based facial detection algorithm, in parallel, on batches of images versus the performance of running the GPU-based facial detection algorithm serially on a sequence of single images. We vary the number of images in which faces are to-be detected from 128 images to 1024 images. For each of the prior mentioned image sets, we measure the performance when the number of images per detection batch is varied from 1 image per batch (sequential) to 1024 images per batch in multiples of 2. We find that the technique of batch image processing improves the performance of a face detection application by approximately 10-11x on to-be-detected image sets consisting of 128, 256, and 1024 images. This performance improvement is attributed to a reduction in the communication overhead between the host CPU and GPU occurring on the PCI bus. Moreover, the technique of batch image processing enables the utilization of more of the GPU's resources that are left underutilized when GPU-based facial detection is done serially on a sequence of single images.
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 | 0.001 | 0.004 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.001 | 0.002 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.007 | 0.003 |
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".