Machine Learning for Images - what's specific, various implementation of image recognition, what can be build using image recognition via machine learning. Image recognition frameworks, CNN, cuda-convnet2, coffee, torch7, theano, etc.
6. 6
pixels
lots of data themselves
in spatial relationships
multiple levels and scales of interest
from low-level features such as texture to high-level
features such as composition
needs data-augmentation
compensate for sensitivity - training with blurred,
cropped, scaled, noised, etc
What is specific: ML for images
7. 7
huge architecture
(both deep and wide) - requires massive amount of
memory and processing power
inter- and intra-class variety
need to describe the universe (huge and diverse
datasets required to feed the data greedy CNNs)
takes time
10+ days for large architectures, even after 10x
reduction thanks to using GPUs
What is specific: ML for images
8. 8
cuda-convnet
python interface, fermi-generation nVidia GPU, no multi-
GPU support
cuda-convnet2
an upgrade to cuda-convnet, optimized for new kepler-
generation nVidia GPUs, multi-GPU support
caffe
deep learning framework, developed by Berkeley Vision
and Learning Center, big community of contributors
Convnet implementations for images
9. 9
torch7
ML algorithms, CNN extensions: fbcnn by Facebook,
used by Google DeepMind
theano
python library, open-ended in terms of network
architecture & transfer functions
…many others
Convnet implementations for images
13. 13
semantic expansion
‘car’ -> ‘vehicle’ -> ‘mean of transportation’
feedback-loop
instant learning from user feedback, to be released in
May
custom training
with specific set of tags
Imagga Auto Tagging
15. 15
Personal Photo Applications
•Apps for mobile photos organization
•Integration in telecom solutions
•Cloud services for consumers
•Device manufacturers
http://getsliki.com
19. 19
Big image data management and
organization
Image driven platforms (DAMs)
Contextual advertising
Interactive/behavioural campaigns (adSense like)
User profiling
insight market, profiling based on image content
Interactive campaigns for brands
new ways to interact with customers
Use Cases