Robust to outlier image inpainting for interface detection in primary separation vessel
Bibliographic record
Abstract
The Primary Separation Vessel (PSV) is integral to the bitumen extraction process in the oil sands industry. Effective control of the interface between the froth and middlings layers is critical for the PSV’s optimal operation. Computer vision techniques can monitor this interface using the images captured from the PSV sight glass. However, image-based models suffer from a lower performance when the image quality is inferior. This is evident in the case of the image data being affected by external degradations. Image inpainting addresses the task of removing unwanted objects and improving the quality of the images. Variational Autoencoder (VAE) can be trained to remove the degradations and restore the image quality. However, in the latent space of a standard VAE which uses a Gaussian distribution for the prior, the input information is spread across all the latent dimensions. This is not suitable particularly in scenarios where the input data consists of a limited number of salient features, without involving complex patterns. This under-regularization of latent space may impact the performance of inpainting when outliers are present in the training data. In this article, a Laplace VAE framework is proposed where the prior is modeled as a Laplace distribution to achieve a better regularization of the latent space and enhance robustness to the outliers in training data. Further, we demonstrate that the Laplace prior promotes sparsity in the latent representations, when there are limited features of interest in the input. This model is used to restore degraded images from a pilot-scale PSV and the interface level is predicted from the restored images using a region-based segmentation method. • Laplace VAE model for robust image inpainting in the presence of outliers. • The model removes occlusions from degraded lab-scale PSV images effectively. • Laplace prior to handle outliers by accommodating the extreme data. • Induces sparsity and improves regularization in the latent space. • Robustness to outliers in training data, compared to Gaussian VAE model.
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 | 0.002 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.000 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
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".