Bibliographic record
Abstract
Anomaly detection is a critical aspect of ensuring product quality and minimising defects in manufacturing processes. The MVTec anomaly detection (MVTec AD) dataset is a well-known benchmark for evaluating the effectiveness of anomaly detection methods in real-world scenarios. In this paper, we present a novel approach to anomaly detection on the MVTec AD dataset by leveraging upon the two-level vector quantised variational autoencoder (VQ-VAE-2) architecture. It encodes defect-free images onto a discrete latent space, and a powerful PixelSnail prior is fitted over the discrete latent space induced by the data. Latent codes with a cross-entropy loss above a certain threshold are assumed to correspond to anomalies. The threshold is usually manually tuned and fixed across the various object and texture categories of the dataset. This is time consuming and suboptimal as a different threshold may be required for each category. We introduce an automatic way of determining this threshold: since the cross-entropy losses follow a log-normal distribution where the distribution for defect-free images lies within the distribution for defect images, we found that a threshold corresponding to half of the maximum loss for defect-free images works well. During inference, the PixelSnail prior is repeatedly called as each pixel is conditioned on the previous pixels in a raster scan order (left to right, top to bottom) which is computationally expensive. We found that the model can be called once for each row of the latent map achieving an order-of-magnitude speedup without a significant drop in performance. Lastly, we show that there is a statistical improvement over the original VQ-VAE and performance is similar to the state-of-the-art.
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) | 0.002 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.003 | 0.001 |
| Research integrity | 0.002 | 0.002 |
| Insufficient payload (model declined to judge) | 0.004 | 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".