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