Computer Vision

30 problems

• CLIP InfoNCE LossMedium
• The classifier head — FC + softmaxMedium
• Feature volume — a conv layer's output shapeEasy
• Convolution — feature-map sizeEasy
• Convolution — one patch stepEasy
• Why CNNs — conv vs dense parameter countMedium
• Convolution — slide a 1D filter (weight sharing)Medium
• Feature map — bias and ReLUEasy
• Flatten an image into a feature vectorEasy
• Global average poolingEasy
• Edges — the horizontal gradientEasy
• Images as tensors — the shapeEasy
• Edge models — magnitude pruningEasy
• Image preprocessing — scale and standardizeMedium
• How many raw pixel values?Easy
• Hierarchy — the receptive field grows with depthEasy
• RGB to grayscale (luminosity)Easy
• Sobel edge magnitudeMedium
• Which filter fires? (template detection)Easy
• Regression vs classification outputEasy
• Edge models — measure sparsityEasy
• CLIP — zero-shot classificationMedium
• Dice lossEasy
• Group / Time-Series CV SplitEasy
• Instance normalizationMedium
• IoU of bounding boxesEasy
• Jaccard similarityEasy
• mAP for Object DetectionHard
• Non-Max Suppression (NMS)Medium
• ViT Patch EmbeddingMedium