// ◉ Multimodal & Vision
Connectionist Temporal Classification (CTC)
Trains speech and handwriting recognizers without frame-level alignment.
CTC 序列对齐损失 2006 Intermediate core
IN PLAIN WORDS
Hundreds of audio frames map to a dozen characters, and nobody knows which frame is which. CTC adds a blank symbol and sums the probability of every alignment that collapses to the correct text, computed efficiently by dynamic programming – so (audio, transcript) pairs alone suffice for end-to-end training.
Used in: Speech recognition (e.g. DeepSpeech 2, wav2vec 2.0 fine-tuning), handwriting and scene-text OCR
01Original paper
Connectionist temporal classification: labelling unsegmented sequence data with recurrent neural networks
Alex Graves, Santiago Fernández, Faustino Gomez, Jürgen Schmidhuber · ICML 2006 · 2006
DOI ↗ ✓ Verified · Crossref API doi:10.1145/1143844.1143891
02Where it sits
Learn the left column first; the right column builds on this one.
Learn first
Then you can learn
- Nothing in the atlas builds on it yet
Related:Whisperwav2vec 2.0
Left: prerequisites · right: builds on this · dashed: related. Full map →
03Learning resources
Official docs, explainers, courses and reference code. All links checked.
04Where the course teaches it
For AI / data
- Verification: verified, checked 2026-10-22.
- JSON: /api/learn/algorithms.json (id:
ctc)