##VGG19 model for Keras
This is the Keras model of the 19-layer network used by the VGG team in the ILSVRC-2014 competition.
It has been obtained by directly converting the Caffe model provived by the authors.
Details about the network architecture can be found in the following arXiv paper:
Very Deep Convolutional Networks for Large-Scale Image Recognition
K. Simonyan, A. Zisserman
Code for Keras plays catch blog post
python qlearn.py
| """ | |
| Possibly correct implementation of an all conv neural network using a single residual module | |
| This code was written for instruction purposes and no attempt to get the best results were made. | |
| References: | |
| Deep Residual Learning for Image Recognition: http://arxiv.org/pdf/1512.03385v1.pdf | |
| STRIVING FOR SIMPLICITY, THE ALL CONVOLUTIONAL NET: http://arxiv.org/pdf/1412.6806v3.pdf | |
| A video walking through the code and main ideas: https://youtu.be/-N_zlfKo4Ec |
| from keras.models import Sequential | |
| from keras.layers.recurrent import Recurrent, GRU, LSTM | |
| from keras import backend as K | |
| # from seya.utils import rnn_states | |
| tol = 1e-4 | |
| def _wta(X): | |
| M = K.max(X, axis=-1, keepdims=True) |
| # VERSION: 0.1 | |
| # DESCRIPTION: PLE | |
| # AUTHOR: Eder Santana <[email protected]> | |
| # COMMENTS: | |
| # Pygame learning environment | |
| # SETUP: | |
| # # Download PLE Dockerfile | |
| # wget ... | |
| # | |
| # # Build atom image |
| import os | |
| os.environ["KERAS_BACKEND"] = "tensorflow" | |
| import tensorflow as tf | |
| from keras.engine import Layer, InputSpec | |
| from keras import backend as K, regularizers, constraints, initializations, activations | |
| class Deconv2D(Layer): | |
| def __init__(self, nb_filter, nb_row, nb_col, | |
| init='glorot_uniform', activation='linear', weights=None, | |
| border_mode='valid', subsample=(1, 1), dim_ordering='tf', |
| '''Trains a multi-output deep NN on the MNIST dataset using crossentropy and | |
| policy gradients (REINFORCE). | |
| The goal of this example is twofold: | |
| * Show how to use policy graidents for training | |
| * Show how to use generators with multioutput models | |
| # Policy graidients | |
| This is a Reinforcement Learning technique [1] that trains the model | |
| following the gradient of the logarithm of action taken scaled by the advantage | |
| (reward - baseline) of that action. | |
| # Generators |
| import numpy as np | |
| import opensfm | |
| import cv2 | |
| from opensfm import dataset | |
| import pylab | |
| import scipy.interpolate | |
| import opensfm.features | |
| def project(shot, point): |
| from __future__ import division | |
| import os | |
| import time | |
| from glob import glob | |
| import tensorflow as tf | |
| from six.moves import xrange | |
| from scipy.misc import imresize | |
| from subpixel import PS | |
| from ops import * |