Computer Vision
This subfield of AI lets machines detect objects, patterns, and features in images and video.
RGB values range from this number to this number.
What is 0 to 255?
This phrase highlights how bad data will lead to a bad model.
Flipping, rotating, and changing brightness or contrast are all examples of this.
What is data augmentation?
This value measures how incorrect a model's prediction is versus the correct answer.
CV works by analyzing these tiny dots of color to find shapes, edges, and features.
What are Pixels?
In (R, G, B) format, this is the value for pure red.
What is (255, 0, 0)?
Spreadsheets, databases, and CSV files are examples of this data type.
What is structured data?
This happens when a model memorizes the training data but fails on new data.
What is underfitting?
This CNN layer detects features like edges, corners, and textures using small filters/kernels.
What is the convolutional layer?
This CV model type looks at the whole image and outputs a single label for it.
What is image classification?
Converting to this needs only 1 color channel, which greatly simplifies the tensor.
What is Grayscale?
JSON and XML are examples of this "in-between" data type.
What is semi-structured data?
This happens when a model is too simple to capture the patterns in the data.
What is underfitting?
This activation function turns negatives to 0 and keeps positives - max(0, x).
What is ReLU?
This CV task finds the edges of objects and draws bounding boxes around them.
What is Object Detection?
This "3D array" is used to store image data
What is a tensor?
The typical percentage of a dataset set aside for the training set.
What is 70-80%?
This occurs when some categories have far more examples than others, biasing the model.
What is class imbalance?
This layer shrinks feature maps to ease computation and reduce overfitting.
What is pooling?
This CV task assigns every pixel in a frame to a label, usually grouping multiple objects.
What is Semantic Segmentation?
An image tensor has these three things multiplied together to form its shape.
What are height, width, and channels?
This Google-owned site is the world's largest data science community, with 50,000+ free datasets.
What is kaggle?
This happens when test data sneaks into training, making results look great but fake.
What is data leakage?
At the end of a CNN, this layer connects every neuron to all neurons from the previous layer to make the final decision.
What is Fully Connected (Dense) layer?