Topics discussed:
- (09:15) Command line arguments and argparse
- Learn how to properly use command line arguments
- Resolve common “USAGE” errors that you will encounter
- (20:50) What IDEs do you recommend for computer vision and deep learning development?
- Sublime Text for smaller projects
- PyCharm for larger projects
- For RPi try to use remote development
- (23:20) How can we use OpenCV OCR for newspaper text recognition?
- Learn how to use EAST to automatically detect text in a newspaper
- Learn how to use Tesseract and OpenCV to OCR the newspaper
- (27:20) Hand gesture recognition on the RPi
- Can hand gesture recognition run in real-time on the RPi?
- What types of models should we use?
- (31:20) How can we improve face recognition accuracy?
- Consider your face detector
- Utilize a combination of facial embeddings and machine learning
- (38:30) How can we build a web crawler for images?
- Check out Scrapy
- Useful for building an image search engine
- (41:12) How can I parse object detection annotations?
- Learn to use libraries such as BeautifulSoup and lxml
- (44:50) Producer/consumer with Python
- How can we use a producer/consumer relationship to build a faster, more efficient deep learning inference system?
- (48:10) How can we perform object detection with OpenCV?
- What do the returned values actually mean?
- How can we take them and convert them to bounding box coordinates and labels?
- (53:20) Threads vs. processes
- When do we use threads?
- And how can processes be used to improve program speed?
- (55:20) How can we do object detection with multiple cameras?
- Use multiple cameras but only a single model
- Use producer/consumer relationship to keep queue filled
- Point of diminishing returns — there is a limit to the number of cameras you can attach to a single computer
Charity of the week:
- Beagle Rescue of Southern Maryland
- Amount donated: $450
Office Hours slides:
Leave a Reply