Deep learning-based image recognition for autonomous driving
Customer Challenges:
- It is difficult to directly solve general object recognition tasks from the input image.
- Labor intensive and costly
- Chance of Human Errors was high
Result:
- 50 % increase in accuracy of object detection
- 20% reduction in manual labour cost
CoreView Solution:
- ~ 10K images annotated and labelled
- Advanced data preprocessing for image data
- Object detection, object localization, movement prediction models
- CNN based advanced image recognition algorithms
- Reinforcement learning approach
Scope:
- Heterogeneous quality of images
- Labelling errors
- Lack of Pre-trained algorithms in early 2017s
