Metrics

Accuracy
The degree to which the predicted labels match the actual labels. High accuracy indicates reliable detection of marine mammals.

Precision
The ratio of true positive predictions to the total predicted positives. High precision means fewer false positives.

Recall
The ratio of true positive predictions to the actual positives. High recall means fewer false negatives.

True negative
Correctly identifying the absence of a marine mammal. Confirms the accuracy of Whale Seeker's detection models.

True positive
Correctly identifying the presence of a marine mammal. This indicates effective detection.

Type 1 error (False positive)
A false positive, incorrectly identifying a marine mammal when there is none (i.e. believing there is a whale when it is in fact a log). Whale Seeker's models minimize false positives to ensure accurate monitoring.

Type II error (False negative)
A false negative, failing to identify a marine mammal when there is one (i.e. believing there is no whale present when there is one). Our solutions aim to reduce false negatives to protect marine life.

Data collection & annotation

Annotation
The process of labeling or marking data, such as images, to identify and describe specific features or objects.

Availability bias
A cognitive bias that causes people to overestimate the importance of information that is readily available to them. This can affect data interpretation and decision-making in wildlife management.

Beaufort scale
A scale for measuring wind speed based on observed sea conditions. Used to exclude certain areas due to difficult weather conditions.

Bias 
A systematic error introduced into sampling or testing that skews results. Recognizing and mitigating bias is crucial in developing accurate detection models.

Glare
Bright, reflected light that can obscure visibility in images. Whale Seeker's solutions exclude high glare areas to improve detection accuracy.

Interobserver bias
Variability in data interpretation between different observers. Minimizing this bias is crucial for reliable data collection and analysis.

Marine mammal observer
A trained individual who monitors marine environments, often from ships or platforms, to detect and document the presence of marine mammals. They play a crucial role in ensuring compliance with environmental regulations and protecting marine life during industrial activities.

Monitor
To observe and check the progress or quality of something over a period of time, for Whale Seeker this equates mainly to different marine mammals and birds as well as environmental factors such as land, ice and rock cover. For example, Whale Seeker's real-time detection tools monitor marine mammals to aid in conservation efforts.

Biology

Carbon sequestration
The process of capturing and storing atmospheric carbon dioxide. This can occur naturally, such as in forest and oceans, or through artificial means.

Carbon cycle
The natural circulation of carbon among the atmosphere, oceans, soil, plants and animals.

Carbon sink
A natural or artificial reservoir that absorbs and stores carbon dioxide from the atmosphere such as forests, oceans and soils. They absorb more carbon dioxide than they release, thus reducing greenhouse gas concentrations in the atmosphere.

Whale fall
The event that occurs when a whale dies and its body sinks to the ocean floor. Whale falls play a significant role in marine biodiversity and nutrient cycling. Additionally, they contribute to carbon sequestration by transferring carbon stored in the whale's body to the deep ocean, where it can remain sequestered for long periods, helping to mitigate the impact of greenhouse gases.

Marine megafauna
Large marine species, including whales, dolphins, seals, and large fish. Monitoring these animals is important for biodiversity assessments and environmental impact studies.

Species at risk
Marine mammal species that are in danger of extinction or becoming endangered. Identifying and monitoring these species is vital for conservation efforts and regulatory compliance.

Remote sensing / Geospatial data

Georeference
Relating a photo or map to the exact coordinates at which these were captured. Essential for mapping marine mammals accurately.

Orthorectification
The process of correcting the geometric distortions in images. Ensures that aerial and satellite images accurately reflect the Earth's surface.

Radiometric correction
Adjusting the pixel values in images to correct for sensor noise and atmospheric effects. Enhances the quality of the data for better detection.

Raster
A grid of pixels or cells, each with a value representing information such as image data. Whale Seeker tools process raster data for detection tasks.

Remote sensing
The acquisition of information about an object or phenomenon without making physical contact. Whale Seeker tools utilize remote sensing to detect marine mammals from the air or sea.

Resampling
The process of interpolating pixel values when transforming images to a new coordinate system. Allows to layer data and to georeference.

Artificial intelligence (AI)

Algorithm
A set of rules or steps designed to solve a problem or perform a task. In Whale Seeker solutions, algorithms process image data to detect marine mammals.
 

Artificial Intelligence (AI)

The simulation of human intelligence processes by machines, especially computer systems. Whale Seeker's AI models detect and identify different marine mammal and bird species in various environments.

​

Classification

The process of identifying and categorizing objects within an image. Whale Seeker's solutions extract different characteristics of marine mammal and bird species.

​

Data drift

Changes in the data distribution over time that can affect model performance. Whale Seeker uses continuous monitoring to detect and address data drift.

​

Detection

The action or process of identifying the presence of marine mammals in images. Whale Seeker's detection tools help monitor these species effectively.

​

Human-in-the-loop

A system where a human input is used in the decision process of an AI algorithm. Ensures Whale Seeker's solutions are accurate and reliable.

​

Model

A mathematical representation of a real-world process. In Whale Seeker's context, models are trained to detect the presence of marine mammals.

​

Model architecture 

The design and structure of a machine learning model, including layers and connections. A well-designed architecture improves detection accuracy.

​

Model framework

A set of tools and libraries used to build and deploy machine learning models. Whale Seeker's models use and develop robust frameworks to ensure reliable performance.

 

Overfitting

A modeling error that occurs when a model learns the details and noise in the training data to the extent that it performs poorly on new data. Whale Seeker strives to develop models that excel not only on benchmark datasets but also on specific client datasets.

​

Supervised Learning

A type of machine learning where the model is trained on labeled data. Whale Seeker's detection models use supervised learning to accurately identify marine mammals.

​

Test set

A subset of data used to evaluate the final model's performance. It ensures the model works well on unseen data.

​

Training set

A subset of data used to train a model. It includes input-output pairs that the model learns from.

 

Transferability

The ability of a model to apply what it has learned in one context to different but related contexts. Whale Seeker's base models are designed to be transferable across various environments.

​

Underfitting

A modeling error that occurs when a model is too simple to capture the underlying pattern in the data. Whale Seeker aims to balance complexity to avoid underfitting.

​

Validation set

A subset of data used to tune model parameters. It helps in assessing the model's performance during development.