Image Recognition and Object Detection with MediaPipe on A…
summary
Shows how to run on-device object detection, classification and segmentation on Android using MediaPipe with concrete code steps: add com.google.mediapipe.tasks, load TensorFlow Lite models into assets, build an MPImage, call objectDetector.detect(mpImage) and parse Detection.boundingBox, score and displayName. Compares models by latency and accuracy using SSD MobileNetV2 (fast), EfficientDet-Lite0/Lite2 (more accurate), demonstrates hair vs DeepLabV3 segmentation masks, and explains allow/deny lists, score thresholds and running modes.
AI-written summary · about
chapters (13)
- 0:00MediaPipe on Android
- 3:13What is MediaPipe
- 6:21MediaPipe tasks and graphs
- 10:16Vision tasks overview
- 13:09Setting up object detection
- 20:06Integrating object detector
- 25:29Processing images and results
- 27:17Understanding detection results
- 31:36Image classification
- 36:35Image segmentation
- 42:38Finding models for MediaPipe
- 45:47On-device ML benefits
- 46:51Summary
related talks
40mOn-Device Machine Learning for Android with MediaPipe
42mOn-Device Machine Learning with Google AI Edge
32mAdvanced on-device ML made easy
50mBuild and deploy custom object detection model with TensorFlow Lite
39mMachine Learning on Android Demystified
34mDetecting Pikachu on Android using Tensorflow Object Detection
38mDeep Learning for detection on a phone
27mMachine Learning on the edge using TensorFlow Lite