Deep Learning Models for Vision Tasks on iOS
Download dependencies folder tf
Extract all archives in tf/models and tf/lib
Put tf folder in same directory level as enVision project folder
Build and Run
Press screen to change running model
Tap a data slot below to select, then tap a detection box to snap
Tap a data slot with two fingers to remove last snap
Press a data slot to clear
https://arxiv.org/abs/1506.02640
YOLO 1 tiny (VOC): Best performance on basic classes
YOLO 1 small (VOC): Better accuracy for basic classes
YOLO 1.1 tiny (COCO): Fast on extended classes
YOLO 2 (COCO): Best accuracy on extended classes
YOLO detector + Jetpac feature extractor from snaps + kNN classifier with Euclidean distance
.
https://arxiv.org/abs/1503.03832
Inception-Resnet-v1 (FaceScrub and CASIA-Webface)
Native iOS face detector + FaceNet feature extractor from snaps + kNN classifier with Euclidean distance
.
https://arxiv.org/abs/1512.00567
Inception v3 (ImageNet)
Can run retrained models instead
.
https://github.com/jetpacapp/DeepBeliefSDK
Jetpac network (ImageNet)
DeepBeliefSDK framework
.
Owner: ID Labs L.L.C.
Original Contributor: Muhammad Hilal