RIGHT DATA AT THE RIGHT TIME: FUELING CNN DEVELOPMENT WITH DATASET ENGINEERING
summary
Continental ADAS builds an automated dataset pipeline that stages GPS/timestamp enrichments, runs CNN pre-labeling on every second of recordings, stores artifacts in MongoDB artifact store and high‑performance image storage, and executes ETL on Kubernetes with a GPU cluster. Data comes from an internal continental fleet, is anonymized by a face/plate‑blurring CNN, and active learning scores select top 1,000–10,000 frames for weekly labeling so models can be retrained within bi-weekly sprint cycles.
AI-written summary · about
chapters (9)
- 0:00Intro
- 1:58ADAS Detection Function
- 3:07The product
- 4:50Software 2.0
- 9:19Pre-development phase
- 13:43Base develpment phase
- 18:33Stabilization phase
- 22:00Fine tuning phase
- 27:17Tools and processes
related talks
20mAccelerating AI pipelines w/ NVIDIA on AWS: An autonomous driving case
21mDataEngBytes 2023 - SYD-T2-04 - Suneeta Mall
29mFrom encoding to machine learning a story of managing media data at Netflix
28mData Curation: Transforming Bytes into AI Gold
40mAccelerate autonomous vehicle development
32mData&AI in TomTom: How are are transforming mobility
30mCrash detector
28mNext-Gen railway asset management with...