"When evaluating an object detection model in computer vision, mean average precision is the most commonly cited metric for assessing performance."
A measure of how precise a model is at prediction time. True positives are divided by all positives that have been guessed.
A measure of performance for a prediction system. Recall is used to assess whether a prediction system is guessing enough. True positives are divided by all possible true positives.
An attempt by a model to replicate the ground truth. A prediction usually contains a confidence value for each class.
Seven Tips for Labeling Images for Computer Vision (Roboflow Blog)
The Difference Between Missing and Null Annotations (Roboflow Blog)
Formats available for importing and exporting on Roboflow
This old machine learning adage conveys a salient machine learning point: unless input data is of high quality, model accuracy — even with the best computer vision architectures — will suffer. Understanding what preprocessing and augmentation are at their core enables data scientists to get the most out of their input data.
Preprocessing steps are image changes that are done to all images, like resizing or converting to grayscale. Preprocessing steps will be applied to training, validation, and testing sets.
What is Image Preprocessing? (Video)
Image augmentation steps are image changes designed only to increase dataset size for better performance, like randomly altering brightness or rotation. Image augmentation steps will be applied only to training images.
Bounding box level augmentation generates new training data by only altering the content of a source image’s bounding boxes. In doing so, developers have greater control over creating training data that is more suitable to their problem’s conditions.
Bounding Box Level Augmentations (Roboflow Blog)
Learning Data Augmentation Strategies for Object Detection - Google Research Team (2019)
AWS, Azure Custom Vision, Google Cloud Vision AutoML
AWS & Azure Custom Vision require Third-Party keys to be entered into your project's workspace. These access points must be set up on your target AWS or Azure account to enable integration with Roboflow.
Characteristics: Specific, Achievable, Measurable
What will our model experience "in the wild…?" 🤔
Through the dataset assembly process, our laser focus might miss a whole host of edge cases that we would naturally consider out-of-scope for our model. While out-of-scope instances are intuitive to us, our model has no way of knowing anything beyond the scope of what it has been shown.
Using pre-trained weights to bootstrap your model’s learning. You are “transferring” the knowledge learned on another dataset and then “fine-tuning” it to learn about your new domain.