First of all, thanks for the release! Second: its a pity that the released model (Resnet) has a highly non-standard layer ("Scale", which has a learnable bias parameter). That means that this model cannot be converted to other frameworks (Torch, MXNet, etc).

Would it be possible to release one of your other models as well, such as the Googlenet model?

Thanks!

I've created a simple Dockerfile and guide to be considered for inclusion in this repo. The goal in mind is to lower the entry barrier for users that want to quickly get started and explore this repo without having to install all of the projects dependencies (Caffe).

hi, I install docker and Build a caffe docker image sucessfully, but is run the programm, it report the error. please help me.

root@iZ25q8sqgi7Z:~/open_nsfw# docker run caffe:cpu caffe --version libdc1394 error: Failed to initialize libdc1394 caffe version 1.0.0-rc3

root@iZ25q8sqgi7Z:~/open_nsfw# docker run --volume=$(pwd):/workspace caffe:cpu \

python ./classify_nsfw.py
--model_def nsfw_model/deploy.prototxt
--pretrained_model nsfw_model/resnet_50_1by2_nsfw.caffemodel
test_image.jpg libdc1394 error: Failed to initialize libdc1394 Traceback (most recent call last): File "./classify_nsfw.py", line 16, in import caffe File "/opt/caffe/python/caffe/init.py", line 1, in from .pycaffe import Net, SGDSolver, NesterovSolver, AdaGradSolver, RMSPropSolver, AdaDeltaSolver, AdamSolver File "/opt/caffe/python/caffe/pycaffe.py", line 15, in import caffe.io File "/opt/caffe/python/caffe/io.py", line 2, in import skimage.io ImportError: No module named skimage.io

Hello. Thank you for sharing this project. The readme suggests fine-tuning for improved performance, and I'm writing to request direction on how to convert the deploy.prototxt file into a train_val.prototxt file for fine-tuning. Or, more simply, to request a sample train_val.prototxt file.

Thank you.

I would like to share how to run NSFT with docker

1. Build a caffe docker image (GPU)

$ cd open_nsfw
$ wget https://raw.githubusercontent.com/BVLC/caffe/master/docker/standalone/gpu/Dockerfile
$ docker build -t caffe:cpu ./

To build for cpu only get the Dockerfile for cpu from Caffe repo

$ wget https://raw.githubusercontent.com/BVLC/caffe/master/docker/standalone/cpu/Dockerfile

2. When the image it's ready, we check caffe installation

$ docker run -ti caffe:cpu caffe --version
caffe version 1.0.0-rc3

3. We run bash in this folder as mapped volume with the option --volume in order to run the training manually

docker run -ti --volume=$(pwd):/workspace caffe:cpu bash

so we are in the docker caffe image

root@8855e6188183:/workspace#

4. Run nsfw in the current container volume

root@8855e6188183:/workspace#python ./classify_nsfw.py --model_def nsfw_model/deploy.prototxt --pretrained_model nsfw_model/resnet_50_1by2_nsfw.caffemodel anderson_210.jpg

and we are done

NSFW score:   0.14057905972

$> docker build -t caffe:cpu https://raw.githubusercontent.com/BVLC/caffe/master/docker/cpu/Dockerfile
[...]
Get:469 http://archive.ubuntu.com/ubuntu xenial-updates/universe amd64 libopencv2.4-jni amd64 2.4.9.1+dfsg-1.5ubuntu1.1 [148 kB]
Get:470 http://archive.ubuntu.com/ubuntu xenial-updates/universe amd64 libopencv2.4-java all 2.4.9.1+dfsg-1.5ubuntu1.1 [433 kB]
Get:471 http://archive.ubuntu.com/ubuntu xenial-updates/universe amd64 libopencv-dev amd64 2.4.9.1+dfsg-1.5ubuntu1.1 [169 kB]
Get:472 http://archive.ubuntu.com/ubuntu xenial/main amd64 libprotobuf-lite9v5 amd64 2.6.1-1.3 [58.4 kB]
Get:473 http://archive.ubuntu.com/ubuntu xenial/main amd64 libprotobuf9v5 amd64 2.6.1-1.3 [326 kB]
Get:474 http://archive.ubuntu.com/ubuntu xenial/main amd64 libprotoc9v5 amd64 2.6.1-1.3 [273 kB]
Get:475 http://archive.ubuntu.com/ubuntu xenial/main amd64 python-decorator all 4.0.6-1 [9326 B]
Get:476 http://archive.ubuntu.com/ubuntu xenial/main amd64 python-numpy amd64 1:1.11.0-1ubuntu1 [1763 kB]
Get:477 http://archive.ubuntu.com/ubuntu xenial-updates/universe amd64 python-pip-whl all 8.1.1-2ubuntu0.4 [1110 kB]
Get:478 http://archive.ubuntu.com/ubuntu xenial-updates/universe amd64 python-pip all 8.1.1-2ubuntu0.4 [144 kB]
Get:479 http://archive.ubuntu.com/ubuntu xenial/main amd64 python-pkg-resources all 20.7.0-1 [108 kB]
Get:480 http://archive.ubuntu.com/ubuntu xenial/main amd64 python-setuptools all 20.7.0-1 [169 kB]
Get:481 http://archive.ubuntu.com/ubuntu xenial/main amd64 libprotobuf-dev amd64 2.6.1-1.3 [473 kB]
Get:482 http://archive.ubuntu.com/ubuntu xenial/main amd64 libsnappy-dev amd64 1.1.3-2 [24.0 kB]
Get:483 http://archive.ubuntu.com/ubuntu xenial/main amd64 protobuf-compiler amd64 2.6.1-1.3 [20.4 kB]
Get:484 http://archive.ubuntu.com/ubuntu xenial/universe amd64 python-scipy amd64 0.17.0-1 [8733 kB]
debconf: delaying package configuration, since apt-utils is not installed
Fetched 234 MB in 10s (23.3 MB/s)
(Reading database ... 4768 files and directories currently installed.)
Preparing to unpack .../perl-base_5.22.1-9ubuntu0.6_amd64.deb ...
Unpacking perl-base (5.22.1-9ubuntu0.6) over (5.22.1-9ubuntu0.5) ...
dpkg: error processing archive /var/cache/apt/archives/perl-base_5.22.1-9ubuntu0.6_amd64.deb (--unpack):
 unable to make backup link of './usr/share/lintian/overrides/perl-base' before installing new version: Invalid cross-device link
dpkg-deb: error: subprocess paste was killed by signal (Broken pipe)
dpkg: error: error creating new backup file '/var/lib/dpkg/status-old': Invalid cross-device link
E: Sub-process /usr/bin/dpkg returned an error code (2)
The command '/bin/sh -c apt-get update && apt-get install -y --no-install-recommends         build-essential         cmake         git         wget         libatlas-base-dev         libboost-all-dev         libgflags-dev         libgoogle-glog-dev         libhdf5-serial-dev         libleveldb-dev         liblmdb-dev         libopencv-dev         libprotobuf-dev         libsnappy-dev         protobuf-compiler         python-dev         python-numpy         python-pip         python-setuptools         python-scipy &&     rm -rf /var/lib/apt/lists/*' returned a non-zero code: 100

Hi guys, I am using open_nsfw to classify images, but some images like the attachment(a avatar's wall, open_nsfw score:0.993217) is marked as nsfw. Anyone knows how to solve this? Thank you very much!

when I try run open_nsfw in multi-process , a runtimeError raised: Traceback (most recent call last): File "./verify_concurrency.py", line 152, in main(sys.argv) File "./verify_concurrency.py", line 149, in main print res.get()
File "/usr/lib/python2.7/multiprocessing/pool.py", line 567, in get raise self._value RuntimeError: Pickling of "caffe._caffe.Net" instances is not enabled (http://www.boost.org/libs/python/doc/v2/pickle.html)

could some expert help check where is wrong as i'm rookie my command is : ./verify_concurrency.py --model_def ./nsfw_model/deploy.prototxt --pretrained_model ./nsfw_model/resnet_50_1by2_nsfw.caffemodel /home/harold/nsfw/pic/1

below is my script verify_concurrency.py : #!/usr/bin/env python """ Copyright 2016 Yahoo Inc. Licensed under the terms of the 2 clause BSD license. Please see LICENSE file in the project root for terms. """

import numpy as np import os import sys import argparse import glob import time from PIL import Image from StringIO import StringIO import caffe from multiprocessing import Pool

def resize_image(data, sz=(256, 256)): """ Resize image. Please use this resize logic for best results instead of the caffe, since it was used to generate training dataset :param str data: The image data :param sz tuple: The resized image dimensions :returns bytearray: A byte array with the resized image """ img_data = str(data) im = Image.open(StringIO(img_data)) if im.mode != "RGB": im = im.convert('RGB') imr = im.resize(sz, resample=Image.BILINEAR) fh_im = StringIO() imr.save(fh_im, format='JPEG') fh_im.seek(0) return bytearray(fh_im.read())

def caffe_preprocess_and_compute(file_name, caffe_transformer=None, caffe_net=None, output_layers=None): """ Run a Caffe network on an input image after preprocessing it to prepare it for Caffe. :param file name file_name: PIL image to be input into Caffe. :param caffe.Net caffe_net: A Caffe network with which to process pimg afrer preprocessing. :param list output_layers: A list of the names of the layers from caffe_net whose outputs are to to be returned. If this is None, the default outputs for the network are returned. :return: Returns the requested outputs from the Caffe net. """ if caffe_net is not None:

    # Grab the default output names if none were requested specifically.
    if output_layers is None:
        output_layers = caffe_net.outputs
    pimg = open(file_name).read()
    start_time = time.time()
    img_data_rs = resize_image(pimg, sz=(256, 256))
    image = caffe.io.load_image(StringIO(img_data_rs))

    H, W, _ = image.shape
    _, _, h, w = caffe_net.blobs['data'].data.shape
    h_off = max((H - h) / 2, 0)
    w_off = max((W - w) / 2, 0)
    crop = image[h_off:h_off + h, w_off:w_off + w, :]
    transformed_image = caffe_transformer.preprocess('data', crop)
    transformed_image.shape = (1,) + transformed_image.shape

    input_name = caffe_net.inputs[0]
    all_outputs = caffe_net.forward_all(blobs=output_layers,
                **{input_name: transformed_image})

    outputs = all_outputs[output_layers[0]][0].astype(float)
    
    return outputs
else:
    return []

end_time = time.time()
print("deal image:%s;start time:%.3f;end time:%.3f;score:%.3f;cost time:%.3f" % (file_name,start_time, end_time , outputs[1] , (end_time - start_time)) )

def main(argv): pycaffe_dir = os.path.dirname(file)

parser = argparse.ArgumentParser()
# Required arguments: input file.
parser.add_argument(
    "input_dir",
    help="Directory of the input image file"
)

# Optional arguments.
parser.add_argument(
    "--model_def",
    help="Model definition file."
)
parser.add_argument(
    "--pretrained_model",
    help="Trained model weights file."
)

args = parser.parse_args()
# Pre-load caffe model.
nsfw_net = caffe.Net(args.model_def,  # pylint: disable=invalid-name
    args.pretrained_model, caffe.TEST)

# Load transformer
# Note that the parameters are hard-coded for best results
caffe_transformer = caffe.io.Transformer({'data': nsfw_net.blobs['data'].data.shape})
caffe_transformer.set_transpose('data', (2, 0, 1))  # move image channels to outermost
caffe_transformer.set_mean('data', np.array([104, 117, 123]))  # subtract the dataset-mean value in each channel
caffe_transformer.set_raw_scale('data', 255)  # rescale from [0, 1] to [0, 255]
caffe_transformer.set_channel_swap('data', (2, 1, 0))  # swap channels from RGB to BGR

start_time  = time.time()
print("The process starttime: %.3f" % start_time) 
file_list = os.listdir(args.input_dir)
multiprocess_pool = Pool(10)
result = []
for file in file_list:

    # Classify.

    p = args.input_dir + '/' + file
    #print("The image %s deal startat %.3f" % (p, time.time()) )
    result.append(multiprocess_pool.apply_async(caffe_preprocess_and_compute, args=(p, caffe_transformer, nsfw_net, ['prob'] ,)))
    #scores = caffe_preprocess_and_compute(p, caffe_transformer=caffe_transformer, caffe_net=nsfw_net, output_layers=['prob'])

    # Scores is the array containing SFW / NSFW image probabilities
    # scores[1] indicates the NSFW probability
    #print "NSFW score:  " , scores[1]
    
    #print("The image %s got score: %.3f at %.3f" % (p, scores[1] ,time.time()) )

print 'Waiting for all subprocesses done...'
multiprocess_pool.close()
multiprocess_pool.join()
print 'All subprocesses done.'
end_time  = time.time()  
print("The process endtime: %.3f" % end_time) 

print("The process cost time is %.3f for %d images" % ( (end_time - start_time), len(file_list)) )
for res in result:
    print res.get()    

if name == 'main': main(sys.argv)

After changing some lines, it is still not working (however the includes and the syntax seem to be ok):

  File "classify_nsfw.py", line 128, in <module>
    main(sys.argv)
  File "classify_nsfw.py", line 104, in main
    image_data = open(args.input_file).read()
  File "/usr/lib/python3.5/codecs.py", line 321, in decode
    (result, consumed) = self._buffer_decode(data, self.errors, final)
UnicodeDecodeError: 'utf-8' codec can't decode byte 0xff in position 0: invalid start byte

Changed

from StringIO import StringIO

to

from io import StringIO

and

print "NSFW score:  " , scores[1]

to

print("NSFW score:  %f" %  float(scores[1]))

I got a build error (using the new URL from #51)

docker build -t caffe:cpu https://raw.githubusercontent.com/BVLC/caffe/master/docker/cpu/Dockerfile

Downloading build context from remote url: https://raw.githubusercontent.com/BVLC/caffe/master/docker/cpu/Dockerfile  1.349kB
Sending build context to Docker daemon  3.072kB
Step 1/12 : FROM ubuntu:16.04
 ---> 0b1edfbffd27
Step 2/12 : LABEL maintainer [email protected]
 ---> Using cache
 ---> d8b42b3a353c
Step 3/12 : RUN apt-get update && apt-get install -y --no-install-recommends         build-essential         cmake         git         wget         libatlas-base-dev         libboost-all-dev         libgflags-dev         libgoogle-glog-dev         libhdf5-serial-dev         libleveldb-dev         liblmdb-dev         libopencv-dev         libprotobuf-dev         libsnappy-dev         protobuf-compiler         python-dev         python-numpy         python-pip         python-setuptools         python-scipy &&     rm -rf /var/lib/apt/lists/*
 ---> Using cache
 ---> 2fb926bad41f
Step 4/12 : ENV CAFFE_ROOT=/opt/caffe
 ---> Using cache
 ---> 7538610c87fa
Step 5/12 : WORKDIR $CAFFE_ROOT
 ---> Using cache
 ---> 46d10669c5cf
Step 6/12 : ENV CLONE_TAG=1.0
 ---> Using cache
 ---> 302ad52eb00c
Step 7/12 : RUN git clone -b ${CLONE_TAG} --depth 1 https://github.com/BVLC/caffe.git . &&     pip install --upgrade pip &&     cd python && for req in $(cat requirements.txt) pydot; do pip install $req; done && cd .. &&     mkdir build && cd build &&     cmake -DCPU_ONLY=1 .. &&     make -j"$(nproc)"
 ---> Running in 4ed26361a7c6
Cloning into '.'...
Note: checking out 'eeebdab16155d34ff8f5f42137da7df4d1c7eab0'.

You are in 'detached HEAD' state. You can look around, make experimental
changes and commit them, and you can discard any commits you make in this
state without impacting any branches by performing another checkout.

If you want to create a new branch to retain commits you create, you may
do so (now or later) by using -b with the checkout command again. Example:

  git checkout -b <new-branch-name>

Collecting pip
  Downloading https://files.pythonhosted.org/packages/0f/74/ecd13431bcc456ed390b44c8a6e917c1820365cbebcb6a8974d1cd045ab4/pip-10.0.1-py2.py3-none-any.whl (1.3MB)
Installing collected packages: pip
  Found existing installation: pip 8.1.1
    Not uninstalling pip at /usr/lib/python2.7/dist-packages, outside environment /usr
Successfully installed pip-10.0.1
Traceback (most recent call last):
  File "/usr/bin/pip", line 9, in <module>
    from pip import main
ImportError: cannot import name main
Traceback (most recent call last):
  File "/usr/bin/pip", line 9, in <module>
    from pip import main
ImportError: cannot import name main
Traceback (most recent call last):
  File "/usr/bin/pip", line 9, in <module>
    from pip import main
ImportError: cannot import name main
Traceback (most recent call last):
  File "/usr/bin/pip", line 9, in <module>
    from pip import main
ImportError: cannot import name main
Traceback (most recent call last):
  File "/usr/bin/pip", line 9, in <module>
    from pip import main
ImportError: cannot import name main
Traceback (most recent call last):
  File "/usr/bin/pip", line 9, in <module>
    from pip import main
ImportError: cannot import name main
Traceback (most recent call last):
  File "/usr/bin/pip", line 9, in <module>
    from pip import main
ImportError: cannot import name main
Traceback (most recent call last):
  File "/usr/bin/pip", line 9, in <module>
    from pip import main
ImportError: cannot import name main
Traceback (most recent call last):
  File "/usr/bin/pip", line 9, in <module>
    from pip import main
ImportError: cannot import name main
Traceback (most recent call last):
  File "/usr/bin/pip", line 9, in <module>
    from pip import main
ImportError: cannot import name main
Traceback (most recent call last):
  File "/usr/bin/pip", line 9, in <module>
    from pip import main
ImportError: cannot import name main
Traceback (most recent call last):
  File "/usr/bin/pip", line 9, in <module>
    from pip import main
ImportError: cannot import name main
Traceback (most recent call last):
  File "/usr/bin/pip", line 9, in <module>
    from pip import main
ImportError: cannot import name main
Traceback (most recent call last):
  File "/usr/bin/pip", line 9, in <module>
    from pip import main
ImportError: cannot import name main
Traceback (most recent call last):
  File "/usr/bin/pip", line 9, in <module>
    from pip import main
ImportError: cannot import name main
Traceback (most recent call last):
  File "/usr/bin/pip", line 9, in <module>
    from pip import main
ImportError: cannot import name main
Traceback (most recent call last):
  File "/usr/bin/pip", line 9, in <module>
    from pip import main
ImportError: cannot import name main
Traceback (most recent call last):
  File "/usr/bin/pip", line 9, in <module>
    from pip import main
ImportError: cannot import name main
The command '/bin/sh -c git clone -b ${CLONE_TAG} --depth 1 https://github.com/BVLC/caffe.git . &&     pip install --upgrade pip &&     cd python && for req in $(cat requirements.txt) pydot; do pip install $req; done && cd .. &&     mkdir build && cd build &&     cmake -DCPU_ONLY=1 .. &&     make -j"$(nproc)"' returned a non-zero code: 1

My PC gets rebooted automatically.

I notice that caffe model was loaded well, but it crashed at scores = caffe_preprocess_and_compute(image_data, caffe_transformer=caffe_transformer, caffe_net=nsfw_net, output_layers=['prob']).

It still reboots when I switch to caffe.set_mode_cpu().

Is there anything that I can do? Thank you.

Running

docker build -t caffe:cpu https://raw.githubusercontent.com/BVLC/caffe/master/docker/cpu/Dockerfile

get

`Step 7/12 : RUN git clone -b ${CLONE_TAG} --depth 1 https://github.com/BVLC/caffe.git . && pip install --upgrade pip && cd python && for req in $(cat requirements.txt) pydot; do pip install $req; done && cd .. && mkdir build && cd build && cmake -DCPU_ONLY=1 .. && make -j"$(nproc)" ---> Running in 0a91bffbd189 Cloning into '.'... Note: checking out 'eeebdab16155d34ff8f5f42137da7df4d1c7eab0'.

You are in 'detached HEAD' state. You can look around, make experimental changes and commit them, and you can discard any commits you make in this state without impacting any branches by performing another checkout.

If you want to create a new branch to retain commits you create, you may do so (now or later) by using -b with the checkout command again. Example:

git checkout -b `

Hi , while running : docker build -t caffe:cpu https://raw.githubusercontent.com/BVLC/caffe/master/docker/cpu/Dockerfile This command , i am getting error at your step 7 , when you are updating pip version.

File "/usr/bin/pip", line 9, in from pip import main ImportError: cannot import name main

As i have already installed latest pip version but when you update in it's giving error because of pip.

Hi Yahoo people.

Your blog post about the release of this model asks that if people develop their own model, to share it here in a bug ticket.

In the same blog post, you mention that you built a squeezenet 1.0 model that performed better, but released the ResNet-thin model (based on fine-tuning imagenet models).

I have trained a squeezenet 1.1 model from-scratch and have achieved, in my tests, slightly better overall accuracy than your released model, so I'd like to share it here and hopefully, as the blog post describes, get it included in your readme.

The model can be found here:

https://github.com/TechnikEmpire/NsfwSqueezenet

I achieved, again in my tests, ~99.1% overall accuracy on ~24,000 random test images, ~12K from each class. The best part is, using OpenCV's DNN module along with Intel's now open-source Inference Engine as the back-end, evaluation on the CPU on modern hardware*, including JPEG decoding and preprocessing, can be executed in ~3msec, or ~333fps. This is more than sufficient for real-time video classification on the CPU, and even ~13x realtime. With the same module, your model executes on the same hardware at ~8msec.

A cutoff (described in the project wiki) was determined through a process of averaging positive classification probability.

This work was sponsored in part by CloudVeil Technologies so I'd like to give them a nod there.

Thank you for open sourcing your model. It inspired me to to develop my model, because it was a clear demonstration of the possibilities in tackling this problem with deep learning.

• "Modern Hardware" here is Intel i7-6700 @ 3.4 ghz. Also of note, this model executes at ~30msec on a piddly little Intel Atom processor.

fixed unable to prepare context: unable to download remote context https://raw.githubusercontent.com/BVLC/caffe/master/docker/standalone/cpu/Dockerfile: failed to GET https://raw.githubusercontent.com/BVLC/caffe/master/docker/standalone/cpu/Dockerfile with status 404 Not Found: 404: Not Found

Hi,

Most modern Linux distributions are packaging python-caffe only for Python3 and hence getting this up and running has been a pain for me, it would be super awesome if the sample script could be made Python3 friendly. I will probably tidy up what I have locally here to fix this and raise a PR.

Best Regards,

Rob