ML Platform Podcast Managing Computer Vision Projects with Michal Tadeusiak
Jul 20, 2022
Michal Tadeusiak, a Computer Vision Project Manager, discusses managing computer vision projects for external clients, experimenting with different architectures, the limitations of using AutoML models, and developing an end-to-end solution. He emphasizes the importance of managing client expectations, building team spirit, and establishing baselines for performance evaluation.
AI Snips
Chapters
Transcript
Episode notes
Starting a Computer Vision Project
- Start by understanding the client's business goals and translate them into technical terms.
- Assess the available data and labels early on to determine project feasibility.
Data Assessment in Computer Vision
- Assessing data difficulty is easier in computer vision projects than in NLP or tabular data projects.
- Having a sample of images allows for quick assessment of project complexity.
Tire Manufacturer Case Study
- DeepSense helped a tire manufacturer with quality assurance by designing a specialized labeling tool.
- The project expanded to include 2D and 3D image analysis and integration into the workflow.