Where can I park? Deep Learning is here to help you!
summary
End-to-end prototype uses images processed on-device to avoid privacy and bandwidth issues: Raspberry Pi Zero W captures and runs TensorFlow models locally, collecting about 17,000 images (13k from one camera) and a curated subset of 4,524 for training. MobileNet out-of-the-box gives ~20% detection, so transfer learning with Inception + SSD is applied, reaching a peak validation 98% (overfitting risk). System reports spot states to an IoT Platform; pipeline works but remains a work in progress.
AI-written summary · about
chapters (7)
- 0:00Introduction and background
- 3:05What is Internet of Things
- 7:01Problem with smart parking
- 10:16Hardware and model selection
- 16:00Training and validation process
- 20:19Live demo and platform
- 23:25Q&A session
related talks
46mImage Recognition with Deep Learning From prototype to production
07mFinding parking spots with Custom Vision and IoT
49mEnable IoT with Edge Computing and Machine Learning
57mMachine Learning in the Age of IoT
59mDeep Neural Networks for Video Applications
19mFrom concept to production: design and optimize computer vision applications on Arm
31mMake your IoT even Smarter with Tensorflow Lite to Design the Future of Vertical Farming
10mSeeing the World With AI: How Deep Learning Enables a New Era of Imaging Technologie