fast-rcnn has been deprecated. Please see Detectron, which includes an implementation of Mask R-CNN.

This code base is no longer maintained and exists as a historical artifact to supplement my ICCV 2015 paper. For more recent work that's faster and more accurrate, please see Faster R-CNN (which also includes functionality for training Fast R-CNN).

Fast R-CNN: Fast Region-based Convolutional Networks for object detection

Created by Ross Girshick at Microsoft Research, Redmond.

Introduction

Fast R-CNN is a fast framework for object detection with deep ConvNets. Fast R-CNN - trains state-of-the-art models, like VGG16, 9x faster than traditional R-CNN and 3x faster than SPPnet, - runs 200x faster than R-CNN and 10x faster than SPPnet at test-time, - has a significantly higher mAP on PASCAL VOC than both R-CNN and SPPnet, - and is written in Python and C++/Caffe.

Fast R-CNN was initially described in an arXiv tech report and later published at ICCV 2015.

License

Fast R-CNN is released under the MIT License (refer to the LICENSE file for details).

Citing Fast R-CNN

If you find Fast R-CNN useful in your research, please consider citing:

@inproceedings{girshickICCV15fastrcnn,
    Author = {Ross Girshick},
    Title = {Fast R-CNN},
    Booktitle = {International Conference on Computer Vision ({ICCV})},
    Year = {2015}
}

Contents

1. Requirements: software

2. Requirements: hardware

3. Basic installation

4. Demo

5. Beyond the demo: training and testing

6. Usage

7. Extra downloads

Requirements: software

1. Requirements for Caffe and pycaffe (see: Caffe installation instructions)

   Note: Caffe must be built with support for Python layers!

   # In your Makefile.config, make sure to have this line uncommentedWITH_PYTHON_LAYER := 1

   You can download my Makefile.config for reference.

2. Python packages you might not have: cython, python-opencv, easydict

3. [optional] MATLAB (required for PASCAL VOC evaluation only)

Requirements: hardware

1. For training smaller networks (CaffeNet, VGG_CNN_M_1024) a good GPU (e.g., Titan, K20, K40, ...) with at least 3G of memory suffices

2. For training with VGG16, you'll need a K40 (~11G of memory)

Installation (sufficient for the demo)

1. Clone the Fast R-CNN repository

   # Make sure to clone with --recursivegit clone --recursive https://github.com/rbgirshick/fast-rcnn.git

2. We'll call the directory that you cloned Fast R-CNN into FRCN_ROOT

   Ignore notes 1 and 2 if you followed step 1 above.

   Note 1: If you didn't clone Fast R-CNN with the --recursive flag, then you'll need to manually clone the caffe-fast-rcnn submodule:

   git submodule update --init --recursive

   Note 2: The caffe-fast-rcnn submodule needs to be on the fast-rcnn branch (or equivalent detached state). This will happen automatically if you follow these instructions.

3. Build the Cython modules

   cd $FRCN_ROOT/lib
   make

4. Build Caffe and pycaffe

   cd $FRCN_ROOT/caffe-fast-rcnn# Now follow the Caffe installation instructions here:#   http://caffe.berkeleyvision.org/installation.html# If you're experienced with Caffe and have all of the requirements installed# and your Makefile.config in place, then simply do:make -j8 && make pycaffe

5. Download pre-computed Fast R-CNN detectors

   cd $FRCN_ROOT./data/scripts/fetch_fast_rcnn_models.sh

   This will populate the $FRCN_ROOT/data folder with fast_rcnn_models. See data/README.md for details.

Demo

After successfully completing basic installation, you'll be ready to run the demo.

Python

To run the demo

cd $FRCN_ROOT./tools/demo.py

The demo performs detection using a VGG16 network trained for detection on PASCAL VOC 2007. The object proposals are pre-computed in order to reduce installation requirements.

Note: If the demo crashes Caffe because your GPU doesn't have enough memory, try running the demo with a small network, e.g., ./tools/demo.py --net caffenet or with --net vgg_cnn_m_1024. Or run in CPU mode ./tools/demo.py --cpu. Type ./tools/demo.py -h for usage.

MATLAB

There's also a basic MATLAB demo, though it's missing some minor bells and whistles compared to the Python version.

cd $FRCN_ROOT/matlab
matlab # wait for matlab to start...# At the matlab prompt, run the script:>> fast_rcnn_demo

Fast R-CNN training is implemented in Python only, but test-time detection functionality also exists in MATLAB. See matlab/fast_rcnn_demo.m and matlab/fast_rcnn_im_detect.m for details.

Computing object proposals

The demo uses pre-computed selective search proposals computed with this code. If you'd like to compute proposals on your own images, there are many options. Here are some pointers; if you run into trouble using these resources please direct questions to the respective authors.

1. Selective Search: original matlab code, python wrapper

2. EdgeBoxes: matlab code

3. GOP and LPO: python code

4. MCG: matlab code

5. RIGOR: matlab code

Apologies if I've left your method off this list. Feel free to contact me and ask for it to be included.

Beyond the demo: installation for training and testing models

1. Download the training, validation, test data and VOCdevkit

   wget http://host.robots.ox.ac.uk/pascal/VOC/voc2007/VOCtrainval_06-Nov-2007.tar
   wget http://host.robots.ox.ac.uk/pascal/VOC/voc2007/VOCtest_06-Nov-2007.tar
   wget http://host.robots.ox.ac.uk/pascal/VOC/voc2007/VOCdevkit_08-Jun-2007.tar

2. Extract all of these tars into one directory named VOCdevkit

   tar xvf VOCtrainval_06-Nov-2007.tar
   tar xvf VOCtest_06-Nov-2007.tar
   tar xvf VOCdevkit_08-Jun-2007.tar

3. It should have this basic structure

   $VOCdevkit/                           # development kit$VOCdevkit/VOCcode/                   # VOC utility code$VOCdevkit/VOC2007                    # image sets, annotations, etc.# ... and several other directories ...

4. Create symlinks for the PASCAL VOC dataset

   cd $FRCN_ROOT/data
   ln -s $VOCdevkit VOCdevkit2007

   Using symlinks is a good idea because you will likely want to share the same PASCAL dataset installation between multiple projects.

5. [Optional] follow similar steps to get PASCAL VOC 2010 and 2012

6. Follow the next sections to download pre-computed object proposals and pre-trained ImageNet models

Download pre-computed Selective Search object proposals

Pre-computed selective search boxes can also be downloaded for VOC2007 and VOC2012.

cd $FRCN_ROOT./data/scripts/fetch_selective_search_data.sh

This will populate the $FRCN_ROOT/data folder with selective_selective_data.

Download pre-trained ImageNet models

Pre-trained ImageNet models can be downloaded for the three networks described in the paper: CaffeNet (model S), VGG_CNN_M_1024 (model M), and VGG16 (model L).

cd $FRCN_ROOT./data/scripts/fetch_imagenet_models.sh

These models are all available in the Caffe Model Zoo, but are provided here for your convenience.

Usage

Train a Fast R-CNN detector. For example, train a VGG16 network on VOC 2007 trainval:

./tools/train_net.py --gpu 0 --solver models/VGG16/solver.prototxt 
    --weights data/imagenet_models/VGG16.v2.caffemodel

If you see this error

EnvironmentError: MATLAB command 'matlab' not found. Please add 'matlab' to your PATH.

then you need to make sure the matlab binary is in your $PATH. MATLAB is currently required for PASCAL VOC evaluation.

Test a Fast R-CNN detector. For example, test the VGG 16 network on VOC 2007 test:

./tools/test_net.py --gpu 1 --def models/VGG16/test.prototxt 
    --net output/default/voc_2007_trainval/vgg16_fast_rcnn_iter_40000.caffemodel

Test output is written underneath $FRCN_ROOT/output.

Compress a Fast R-CNN model using truncated SVD on the fully-connected layers:

./tools/compress_net.py --def models/VGG16/test.prototxt 
    --def-svd models/VGG16/compressed/test.prototxt 
    --net output/default/voc_2007_trainval/vgg16_fast_rcnn_iter_40000.caffemodel# Test the model you just compressed./tools/test_net.py --gpu 0 --def models/VGG16/compressed/test.prototxt 
    --net output/default/voc_2007_trainval/vgg16_fast_rcnn_iter_40000_svd_fc6_1024_fc7_256.caffemodel

Experiment scripts

Scripts to reproduce the experiments in the paper (up to stochastic variation) are provided in $FRCN_ROOT/experiments/scripts. Log files for experiments are located in experiments/logs.

Note: Until recently (commit a566e39), the RNG seed for Caffe was not fixed during training. Now it's fixed, unless train_net.py is called with the --rand flag. Results generated before this commit will have some stochastic variation.

Extra downloads

• Experiment logs

• PASCAL VOC test set detections

• voc_2007_test_results_fast_rcnn_caffenet_trained_on_2007_trainval.tgz

• voc_2007_test_results_fast_rcnn_vgg16_trained_on_2007_trainval.tgz

• voc_2007_test_results_fast_rcnn_vgg_cnn_m_1024_trained_on_2007_trainval.tgz

• voc_2012_test_results_fast_rcnn_vgg16_trained_on_2007_trainvaltest_2012_trainval.tgz

• voc_2012_test_results_fast_rcnn_vgg16_trained_on_2012_trainval.tgz

• Fast R-CNN VGG16 model trained on VOC07 train,val,test union with VOC12 train,val