Computer Vision For iOS Developers Course

Discover how computer vision lets computers see and interpret digital images, contrasting it with image processing while examining object classification, detection, and segmentation.

Explore how object detection and semantic segmentation power face detection, people counting, vehicle tracking, speed estimation for ships, self-driving car features, marketing, and sports computer vision.

Explore five image annotation tools to create datasets and label images for training computer vision models, using bounding boxes, polygons, and semantic segmentation with export options.

Explore tools and environments to train mobile computer vision models, including cloud and local training, data preparation, annotations, and model types like object detectors and image classifiers.

MakeML overview and initial setup for beginners, highlighting data labeling, training neural networks, and model training in a few clicks. Create data sets and micro projects to start training now.

Learn to create your first object detection dataset with bounding boxes by importing dog images or video frames, annotating with dog labels and colors, and exporting in multiple formats.

The lecture demonstrates exporting object detection datasets in multiple formats, including coco, Pascal, and coco-like json, with images and bounding-box annotations organized by image, annotation, and category.

Create a polygon semantic segmentation dataset by importing images, drawing polygon annotations with titles and colors, and exporting with augmentation settings to produce grayscale masks.

Explore the dataset store and Google dataset search to locate millions of data sets for object detection, with examples from gun and fish datasets, licenses, and import into projects.

Explore image dataset augmentation for dog images by applying crops, rotations, flips, and color changes to expand from 20 to 2,000 images and export the augmented dataset.

Learn how to adjust training hyperparameters—batch size, iterations, and learning rate—to improve object detection and semantic segmentation models, reduce loss, and optimize model quality.

Prepare a dribble app dataset for soccer ball detection by collecting 182 images with bounding boxes, then apply color and contrast augmentation for training a detector.

Review the dribble dataset trained model and its loss, replace the model in the iOS app, and run it on a device to observe soccer ball detection.

Implement dribble detection in an iOS vision app by processing camera frames, detecting ball centers, counting touches via left-right motion, and updating a green circle overlay with the count.

Explore a PepsiCo case of using computer vision to sort potatoes by weight for iOS developers, detailing three datasets for detection and segmentation and presenting training results.

the video demonstrates training multiple potato detection models using different datasets and augmentations, showing how data quality and augmentation choices influence performance and guiding future iOS app integration.

Learn how to count potato weight by processing camera frames with a model, compute potato area, derive volume via square root, apply a coefficient, and display the weight on screen.