Keras | VGG16 Places365 - VGG16 CNN models pre-trained on Places365-Standard for scene classification

You have just found the Keras models of the pre-trained VGG16 CNNs on Places365-Standard (~1.8 million images from 365 scene categories).

Overview

CNNs trained on Places365 database (latest subset of Places2 Database) could be directly used for scene recognition, while the deep scene features from the higher level layer of CNN could be used as generic features for visual recognition.

Paper

The Keras models has been obtained by directly converting the Caffe models provived by the authors (all the original Caffe-based resources can be found there).

More details about the architecture of the networks can be found in the following paper:

Places: A 10 million Image Database for Scene Recognition
Zhou, B., Lapedriza, A., Khosla, A., Oliva, A., & Torralba, A.
IEEE Transactions on Pattern Analysis and Machine Intelligence

Please consider citing the above paper if you use the pre-trained CNN models.

Contents:

This repository contains code for the following Keras models:

• VGG16-places365
• VGG16-hybrid1365

Usage:

All architectures are compatible with both TensorFlow and Theano, and upon instantiation the models will be built according to the image dimension ordering set in your Keras configuration file at ~/.keras/keras.json. For instance, if you have set image_dim_ordering=tf, then any model loaded from this repository will get built according to the TensorFlow dimension ordering convention, "Width-Height-Depth".

Pre-trained weights can be automatically loaded upon instantiation (weights='places' argument in model constructor for all image models). Weights are automatically downloaded.

Examples

Classify Places classes with VGG16-places365

import os
import urllib2
import numpy as np
from PIL import Image
from cv2 import resize

from vgg16_places_365 import VGG16_Places365

TEST_IMAGE_URL = 'http://places2.csail.mit.edu/imgs/demo/6.jpg'

image = Image.open(urllib2.urlopen(TEST_IMAGE_URL))
image = np.array(image, dtype=np.uint8)
image = resize(image, (224, 224))
image = np.expand_dims(image, 0)

model = VGG16_Places365(weights='places')
predictions_to_return = 5
preds = model.predict(image)[0]
top_preds = np.argsort(preds)[::-1][0:predictions_to_return]

# load the class label
file_name = 'categories_places365.txt'
if not os.access(file_name, os.W_OK):
    synset_url = 'https://raw.githubusercontent.com/csailvision/places365/master/categories_places365.txt'
    os.system('wget ' + synset_url)
classes = list()
with open(file_name) as class_file:
    for line in class_file:
        classes.append(line.strip().split(' ')[0][3:])
classes = tuple(classes)

print('--SCENE CATEGORIES:')
# output the prediction
for i in range(0, 5):
    print(classes[top_preds[i]])

# --PREDICTED SCENE CATEGORIES:
# cafeteria
# food_court
# restaurant_patio
# banquet_hall
# restaurant

Extract features from images with VGG16-hybrid1365

import urllib2
import numpy as np
from PIL import Image
from cv2 import resize

from vgg16_hybrid_places_1365 import VGG16_Hybrid_1365

TEST_IMAGE_URL = 'http://places2.csail.mit.edu/imgs/demo/6.jpg'

image = Image.open(urllib2.urlopen(TEST_IMAGE_URL))
image = np.array(image, dtype=np.uint8)
image = resize(image, (224, 224))
image = np.expand_dims(image, 0)

model = VGG16_Hybrid_1365(weights='places', include_top=False)
features = model.predict(image)

References

• A 10 million Image Database for Scene Recognition - please cite this paper if you use the VGG16-places365 or VGG16-hybrid1365 model in your work.
• Learning Deep Features for Scene Recognition using Places Database

Additionally, don't forget to cite this repo if you use these models:

@misc{gkallia2017keras_places365,
title={Keras-VGG16-Places365},
author={Grigorios Kalliatakis},
year={2017},
publisher={GitHub},
howpublished={\url{https://github.com/GKalliatakis/Keras-VGG16-places365}},
}

Licensing

• All code in this repository is under the MIT license as specified by the LICENSE file.
• The VGG16-places365 and VGG16-hybrid1365 weights were originally ported from the ones released by CSAILVision under the MIT license but resulted in an identical & low confidence predictions issue.
• The current version of the VGG16-places365 and VGG16-hybrid1365 weights are the ones released by landmark-recognition-challenge under the GNU General Public License v3.0

We are always interested in how these models are being used, so if you found them useful or plan to make a release of code based on or using this package, it would be great to hear from you.

Where to get other trained models?

More info on downloading, converting, and submitting other models can be found on the main Keras | Application Zoo repository.

Questions and Comments

If you have any suggestions or bugs to report you can pull a request or start a discussion.