ManySpace
Türkçe Sign in Sign up

AI & MLRecommendedAIML-MM-002

Contrastive Image-Text Pretraining

Aligning image and text encoders on web-scale pairs, enabling zero-shot classification by text prompts.

Difficulty
3 / 5
Importance
4 / 5
Study time
1–2 h
Kind of work
Mixed

Why it matters

The learner can perform zero-shot classification with image-text similarity and explain how prompt wording changes the result.

The route to this topic

Route AIML-MM-002
What Is a Model? Tasks: Regression, Classification, and Beyond Features and Targets Learning Paradigms: Where the Supervision Signal Comes From Classification Framing Linear Regression I: Prediction and Residuals Embeddings Images as Tensors Scores and Logits Self-Supervised Learning Loss and Fitting: Squared Error Similarity in Vector Space Multimodal Representations Multiclass Scores and Softmax Objective Functions and Empirical Risk Query, Key, Value Likelihood Similarity Scores Cross-Entropy Loss Weighted Aggregation Self-Attention Contrastive Learning The Attention Block Vision Transformers Contrastive Image-Text Pretraining