Learn Before
Explain how to address an information gap in an autonomous driving pipeline.
Question: Suppose an autonomous driving pipeline consisting of car detection, pedestrian detection, and path planning performs far below humans, even though each component performs near human-level given its inputs. Explain why this happens, how it can be resolved without making path planning consume raw images, and the role of the additional detector.
Sample answer: This situation indicates that the pipeline is suffering from an information gap, meaning it is missing key information necessary for the overall task. While each component performs well individually on its given inputs, the path planning component lacks critical road information, such as lane markings, to make safe decisions. To resolve this without making the path planner process raw images directly, we can add a lane-marking detector to the pipeline. The lane-marking detector processes camera images and provides structured lane coordinates directly to the path planning component, supplying the missing information while keeping the path planner's inputs clean and structured.
Key points:
- The overall system underperforms due to an information gap (missing information).
- Each component is near human-level given its inputs, but the inputs to the path planner are insufficient.
- Adding a lane-marking detector provides the missing lane information.
- The lane-marking detector supplies this information without requiring the path planner to process raw camera images directly.
Rubric: The response must explain the concept of an information gap where individual components perform well but the overall system lacks necessary inputs. It must identify adding a lane-marking detector as the solution and state that this detector provides structured information to the path planner rather than forcing the path planner to consume raw camera images directly.
0
1
References
Machine Learning Yearning (Deeplearning.ai)
Machine Learning Yearning (Deeplearning.ai)
Machine Learning Yearning (Deeplearning.ai)
Machine Learning Yearning (Deeplearning.ai)
Machine Learning Yearning (Deeplearning.ai)
Machine Learning Yearning (Deeplearning.ai)
Machine Learning Yearning (Deeplearning.ai)
Machine Learning Yearning (Deeplearning.ai)
Machine Learning Yearning (Deeplearning.ai)
Machine Learning Yearning (Deeplearning.ai)
Machine Learning Yearning (Deeplearning.ai)
Machine Learning Yearning (Deeplearning.ai)
Machine Learning Yearning (Deeplearning.ai)
Machine Learning Yearning (Deeplearning.ai)
Machine Learning Yearning (Deeplearning.ai)
Machine Learning Yearning (Deeplearning.ai)
Machine Learning Yearning (Deeplearning.ai)
Machine Learning Yearning (Deeplearning.ai)
Machine Learning Yearning (Deeplearning.ai)
Machine Learning Yearning (Deeplearning.ai)
Machine Learning Yearning (Deeplearning.ai)
Machine Learning Yearning (Deeplearning.ai)
Machine Learning Yearning (Deeplearning.ai)
Machine Learning Yearning (Deeplearning.ai)
Machine Learning Yearning (Deeplearning.ai)
Machine Learning Yearning (Deeplearning.ai)
Machine Learning Yearning (Deeplearning.ai)
Machine Learning Yearning (Deeplearning.ai)
Machine Learning Yearning (Deeplearning.ai)
Machine Learning Yearning (Deeplearning.ai)
Machine Learning Yearning (Deeplearning.ai)
Machine Learning Yearning (Deeplearning.ai)
Machine Learning Yearning (Deeplearning.ai)
Machine Learning Yearning (Deeplearning.ai)
Machine Learning Yearning (Deeplearning.ai)
Machine Learning Yearning (Deeplearning.ai)
Machine Learning Yearning (Deeplearning.ai)
Machine Learning Yearning (Deeplearning.ai)
Machine Learning Yearning (Deeplearning.ai)
Machine Learning Yearning (Deeplearning.ai)
Tags
Machine Learning
Deep Learning
Supervised Learning
Dive into Deep Learning @ D2L
Data Science
Machine Learning Strategy
Related
Autonomous Driving Pipeline Encodes Simpler Subtasks
Why can an autonomous-driving pipeline perform far below human level even when each of its components individually operates near human level?
True or False: In an autonomous-driving pipeline, every component—including path planning—must use a machine-learned model.
Adding a _____ detector to an autonomous-driving pipeline can supply missing road-context information to the path planner without requiring the planner to consume raw camera images.
What are the three components in the simple autonomous driving pipeline described in Machine Learning Yearning?
True or False: Every component in a machine learning pipeline must be a learned model.
When an autonomous driving pipeline underperforms overall despite each component being near human-level, adding a _____ detector can supply the missing information.
Match each autonomous driving pipeline component to its role in the pipeline.
Order the reasoning steps for diagnosing why an autonomous driving pipeline underperforms humans despite each sub-component being individually strong.
If each sub-component of an autonomous driving pipeline is near human-level given its inputs but the full system underperforms humans, what is the most likely cause?
True or False: According to Machine Learning Yearning, obtaining training data for car and pedestrian detectors is relatively easy due to existing datasets and crowdsourcing.
Machine Learning Yearning states that _____ (such as Amazon Mechanical Turk) can be used to obtain larger labeled datasets for training car and pedestrian detectors.
Match each pipeline diagnosis concept to its correct description in the autonomous driving example.
Order the steps for fixing an information gap in the autonomous driving pipeline by adding a lane-marking detector.
Explain how to address an information gap in an autonomous driving pipeline.
Diagnose pipeline failure in an autonomous vehicle system.
Explain the benefit of using a lane-marking detector in a pipeline.