Speeding up data labeling with neural nets: pipeline, metrics and life hacks
summary
Describes building a data-labeling pipeline that uses neural networks to speed up complex annotation tasks, detailing the system architecture, the metrics used to evaluate model-assisted labeling, workload generation, and comparisons between model outputs and human annotators. Discusses practical issues and engineering tips for integrating neural models into the annotation workflow.
AI-written summary · about
chapters (7)
- 0:00Welcome
- 4:16Complex data labeling problems
- 9:33On-the-fly labeling speedup
- 19:50Pre-labeling automation
- 23:14Neural nets as building blocks
- 30:00Economic impact and conclusions
- 31:27Q&A
related talks
40mData labelling to scale your ML and AI pipelines
12mIncreasing the Speed and Accuracy of Data Labeling Through an AI Assisted Interface
29mWhy Deep Learning is driving demand for Human in the Loop
36mProgrammatically creating and managing training data
51mEfficient Deep Learning with Humans in the Loop
21mFrom 0 to a crowd sourced neural network in a jiffy
20mPragmatic Deep Learning for image labelling
37mBest Practices for Building AI Datasets